Comments Off on Deep Dive into the Past, Present, & Future of Income Share Agreements – Erik
Imagine a world where you had a personal board of advisors — the people you most admire and respect — and you gave them upside in your future earnings in exchange for helping you (e.g our good friendMr. Mike Merrill.)
Imagine if there was a “Kickstarter for people” where you could support up-and-coming artists, developers, entrepreneurs — when they need the cash the most, and most importantly, you’d only profit when they profit.
Imagine if you could diversify by pooling 1% of your future income with your ten smartest friends.
Now think about how much you’d go out of your way to help, say, your brother-in-law or step-siblings. Probably much more than a stranger. Why is that?
To pose a thought experiment: If you didn’t know your cousins were related to you, you might treat them like any other person. But because we have this social context of an “extended family,” you have a sort of genetic equity in them — a feeling that your fates are shared and it’s your responsibility to support them.
This begs the question: How can we create the social context needed for people to truly care about others outside of their extended family?
If you believe that markets and trade have helped the world become a less violent place — because why hurt someone when it’ll also take money out of your pocket? — then you should believe that adding more markets (with proper safeguards) will make the world even less violent.
This is the hope of income share agreements (ISAs).
ISAs align economic incentives in ways that encourage us to help others beyond our extended family, give people economic opportunity who don’t have it today, and free people from the shackles of debt.
What are these ISAs you speak of?
An Income Share Agreement is a financial arrangement where an individual or organization provides something of value to a recipient, who, in exchange, agrees to pay back a percentage of their income for a certain period of time.
In the context of education, ISAs are a debt-free alternative to loans.
Rather than go into debt, students receive interest-free funding from an investor or benefactor. In exchange, the student agrees to share a percentage of future income with their counterparty. They come in different shapes and sizes, but almost always with terms that take into account a plethora of potential scenarios.
“Part of the elegance of an ISA is that the lender only wants a share of income when the borrower is getting a regular income “If you’re unemployed or underemployed, they’re not interested… you’re automatically getting a suspension of payments when you’re not doing well.”
There is a long and storied history of income share agreements, but they’ve only recently become popular due to the rise of Lambda School, a school that lets students attend for free and, if they do well after school, pay a percentage of their income until they pay Lambda back.
Wait, a popular meme sarcastically asks, did you just invent taxes?
No. Lambda only gets paid if and only if the student earns a certain amount after graduation. In other words, incentives are aligned. The student is the customer. Not the government. Not the state. Not the parents.
To be sure, it’s early days for ISAs: Adverse selection, legalization, concerns about individuals being corporations (derivatives? Shorting people?!) — there’s a lot left to figure out.
Still, it’s an idea that once you see, you can’t unsee.
Here’s a hypothetical story to help you picture how ISAs work:
Picture Janet, a Senior at Davidson High School. She has a 4.0 GPA, is captain of the debate team and star center forward of the Varsity Soccer team. She’s a shoo-in for a top 20 university, but her parents can’t afford it even with a scholarship, so she’s not even going to apply, and is headed for State. Then she learns from a news article that she’s a pretty good bet as someone who’s going to succeed down the road, and that might allow her to put some much needed cash towards her education. She goes for it, makes a profile on an ISA, and sure enough, a few strangers bet $50,000 on her college education! She immediately gets to work filling out Ivy League scholarship applications.
Throughout college, she keeps in touch with her investors, they give her advice, and because of her interest in politics, one even helps her get an internship with a governor’s election campaign over the summer. Once she graduates, she knows the clock is ticking — at 23 she’ll need to start paying back the investors 5% of her after tax income, so she hustles to work her way through the ranks.
From age 23 to 33, the payback period, Janet becomes a lawyer at a top tier firm, and the investors make a 3x cash on cash return
The above is purely hypothetical.
ISAs for traditional higher education are much more complicated than say, vocational training, where there is more direct alignment of ‘skills-development-to-job’ pathway for students. But, the beauty of ISAs is in their flexibility, so there is lots of room for innovation.
So: this is the dream — why hasn’t it happened yet?
ISAs and other related instances of securitizing human capital have been tried. Here’s a brief history:
Economist Milton Friedman Proposes Use of ISAs in Education —
In a section devoted specifically to vocational and professional education, Friedman proposed that an investor could buy a share in a student’s future earning prospects.
It’s worth noting that the barriers to adoption that Friedman identified back in the 1950s still hold true today:
The potential high costs of administration;
The sheer novelty of the idea;
The reluctance to think of investments in human beings as comparable to investments in physical assets; and
Legal and conventional limitations by suitable financial intermediaries
Society might not have been ready for ISAs in the 1950s, but 16 years later, another Nobel Prize-winning economist, James Tobin, would help launch the first ISA option for college students at Yale University.
Yale experiments with ISAs —
In the 1970s, Yale University ran an experiment called the Tuition Postponement Option (“TPO”). The TPO was a student loan program that enabled groups of undergraduates to pay off loans as a “cohort” by committing a portion of their future annual income.
Students who signed up for the program (3,300 in total) were to pay four percent of their annual income for every $1,000 borrowed until the entire group’s debt had been paid off. High earners could buy out early, paying 150% of what was borrowed plus interest.
Within each cohort, many low earners defaulted, while the highest earners bought out early, leaving a disproportionate debt burden for the remaining graduates.
Administrators also did not account for the changes to the tax code and skyrocketing inflation in the 1980s, which only exacerbated the inequitable arrangement.
“We’re all glad it’s come to an end,” It was an experiment that had good intentions but several design flaws.” — Yale President Richard Levin.
While the TPO is generally considered a failure, it was the first instance of a major university offering ISAs and a useful example for how not to structure ISAs — specifically, pooling students by cohort and allowing the highest earning students to buy out early.
ISAs as a Financial Aid Option —
It would be decades after Yale’s failed experiment before universities started experimenting again with ISAs, but today a company called Vemo Education is leading the way.
This is a crucial point: Vemo isn’t competing directly with loans, but instead is unlocking other sorts of value (i.e., helping students better choose their college). The key here is that Vemo links an individual’s fortunes to the institution’s fortunes. The company helps universities signal value to students by helping them offer ISAs that signal that the university wants to better align cost with value of its higher education program.
The first company that Vemo partnered with to offer ISAs was PurdueUniversity.
In 2016, Purdue University began partnering with Vemo Education to offer students an ISA tuition option through its “Back a Boiler” ISA Fund. They started with a $2 million fund, and since then have raised another $10.2 million and have issued 759 contracts totaling $9.5 million to students.
Purdue markets its ISA offering as an alternative to private student loans and Parent PLUS Loans. Students of any major can get $10,000 per year in ISA funding at rates that vary between 1.73% and 5.00% of their monthly income. Purdue caps payments at 2.5x the ISA amount that students take out and payment is waived for students making less than $20,000 in annual income.
In the last few years, Vemo has emerged as the leading partner for higher education institutions looking to develop, launch and implement ISAs. In 2017, Vemo powered $23M of ISAs for college students across the US.
Upstart: A Short-Lived Attempt at “Kickstarter for People” —
Fintech company Upstart initially launched with a model of “crowdfunding for education”. However, they eventually pivoted to offering traditional loans when they realized that their initial model was simply not viable.
Why? Not enough supply.
The fact that only accredited investors (over $1m in net worth) could invest severely limited the total potential funders on the site. And yet, while Upstart never got enough traction (they pivoted successfully), they paved the way for a platform like it to eventually be built.
ISAs for Vocational Training —
While Upstart failed to gain traction, technical educational bootcamps have seen tremendous growth while offering their students ISAs to finance their education.
And Lambda School is leading the way.
Lambda School is an online bootcamp that trains students to become software engineers at no upfront cost. Instead of paying tuition, students agree to pay 17% of their income for the first two years that they’re employed. Lambda School includes a $50,000 minimum income threshold and caps total payments at an aggregate $30,000. They also give students the option to pay $20,000 upfront if they’d rather not receive an ISA.
Lambda School students enroll for nine months and end up with 1,500–2,000 hours of training, comparable to the level of training they’d receive during a CS-focused portion of a four-year CS degree.
“Lambda School looks like a charity from the outside, but we’re really more like a hedge fund.
We bet that smart, hardworking people are fundamentally undervalued, and we can apply some cash and leverage to fix that, taking a cut.” — Austin Allred (Lambda School CEO)
In our opinion, Lambda is legitimizing ISAs and may just be the wedge that makes ISAs mainstream.
An Outlook for the Future of ISAs
Given where we are today, and with the potential for this type of financial innovation, what might the future look like?
There are three major themes in particular that get us excited for the future of ISAs: aggregation, novel incentive structures, and crypto.
We believe that it’s possible to pool together various segments of people to decrease overall risk of that population and provide more to each individual person.
If we assume that each individual is fairly independent from each other, this should be a possibility. As risk declines, your expected return should increase. And as your expected return increases, more investors and ISA providers will likely jump in to provide even more capital for more people.
“There is no reason you have to do this at the individual level. Most likely, it will first occur in larger aggregated groups — based on either geography, education, or other group characteristics. As with the housing market, it is important to aggregate enough individual sample points to reduce risk.” — Dave McClure
Another take on aggregation could be an individual electing to group together with their close friends or peers.
This can have the magical benefit of further aligning incentives with those around you, increasing the value of cooperation, lowering downside risk, and promoting more potential risk taking or thinking outside the box, all of which should have the benefit of increasing economic growth.
In addition to that, being able to take a more active role in a friend’s life (helping when need be, sharing in their wins, supporting in their losses, etc.) can be an extremely rewarding experience. That said, there are some definite downsides and risks to be aware of with these types of arrangements.
Novel Incentive Structures —
How can we create financial products to incentivize service provides (i.e. teachers, doctors, etc.) where they are indirectly having massive impacts to income from future generations?
Just imagine if every teacher was able to take even just a tiny percentage of every one of their students’ future earnings the difference that tweak could make. Teachers today unfortunately don’t make nearly as much money as they should given the significant consequences they have on future generations. A great teacher can create the spark for the next Einstein or Elon Musk. A terrible teacher could damage the potential Einstein or Elon Musk enough where they never realize their potential. Imagine how many more incredible people we could have.
There will always be incredible teachers regardless of monetary return, but we bet there could be more. It all comes down to aligning incentives.
This same thinking can be applied to other service providers like doctors. Currently, doctors are paid the same amount (all else equal) whether they succeed or not in a life-saving surgery. But what if the service provider also took a tiny fraction of future earnings from their patient? Incentives are more aligned. That doctor may not even realize it, but they likely would work a bit harder knowing what’s at stake.
Crypto can securitize so much more than we currently do; in essence, we could tokenize ourselves and all future income. Once those personal tokens exist, they can be traded instantly anywhere on the world with infinite divisibility. Arbitrageurs and professional traders could create new financial products (i.e. ISA aggregations) and buy / sell with each other to price things to near perfection.
We’d love to continue the conversation! This is a fascinating space with a ton of opportunity. If you’re thinking about or building anything here, feel free to leave your comments or reach out to talk more.
Special shoutout to David Weinstein & Jake Hallac for their help writing as well as Ray Batra, Dani Grant, Zander Adell, Dave McClure, Sam Lessin and Alex Marcus for their help reviewing / editing!
Appendix: Addressing Common Concerns —
Isn’t giving up the legal right to a portion of future income equivalent to modern-day indentured servitude?
Quick refresher: Indentured servants were immigrants who bargained away their labor (and freedom) for four-to-seven years in exchange for passage to the British colonies, room, board and freedom dues (a prearranged severance). Most of these immigrants were English men who came to British colonies in the 17th century.
On the surface this seems like a decent deal, but not so fast. They could be sold, lent out or inherited. Only 40% of indentured servants lived to complete the terms of their contracts. Masters traded laborers as property and disciplined them with impunity, all lawful at the time.
Rebuttal: We are in no way advocating a return to indentured servitude (voluntary or otherwise). Modern-day ISAs must be structured to have proper governance, ensure alignment of interests and contain legal covenants that protect both parties.
We are advocating for ISAs that (i) are voluntary, (ii) do not force the recipient to work for the investor, and (iii) are a promise to share future income, not an obligation to repay a debt.
ISAs are unregulated. How do we structure and enforce ISAs without a legal framework to rely on?
Our Response: ISAs offered by Lambda School, Holberton School and other companies are legal under current US law. To the best of our knowledge, all companies offering ISAs operate according to best practices (i.e., consumer disclosure and borrower protections) as set forth in proposed federal legislation.
The Investing in Student Success Act (H.R.3432, S.268) has been proposed in both the US House of Representatives and the US Senate. Under this legislation, ISAs would be classified as qualified education loans (rather than equity or debt securities), making them dischargeable in bankruptcy. Furthermore, the bill would exempt ISAs from being considered an investment company under the Investment Company Act of 1940.
Importantly, the bill includes consumer protections (i.e., required disclosures, payback periods, payback caps, and limits on income share amounts). The bill also includes tax stipulations that preclude ISA recipients from owning any taxes and limiting taxes for investors to apply to profits earned from ISAs.
Given that ISAs are riskier than student loans, but don’t require the same qualifications, aren’t ISAs prone to adverse selection?
Quick refresher: Adverse selection describes a situation in which one party has information that the other does not have. To fight adverse selection, insurance companies reduce exposure to large claims by limiting coverage of raising premiums.
Our Response: In September 2018, Purdue University published a research study that looked into adverse selection in ISAs. The study concluded that there was no adverse selection by student ability among borrowers. However, ISA providers need to properly structure the ISA so as not to cap a recipient’s upside by too much. In addition, this risk can be mitigated by (i) offering a structured educational curriculum for high-income jobs and (ii) an application process that ensures that students have the ability and motivation to complete a given vocational program.
Our Response: Properly structured ISAs paired with effective offerings (i.e., skills-based training, career development assistance) have the potential to mitigate inequality and discriminatory practices. ISA programs like Lambda School require students to be motivated to succeed and have enough income to complete the program, but in no way discriminate based on age, gender or ethnicity.
However, as ISAs become more common, new legislation must include explicit protections to guard against discrimination in administration of ISAs (especially given that it’s unclear whether the Equal Credit Opportunity Act would apply to ISAs since they aren’t technically loans).
Can’t students simply refuse to pay once they start earning income after graduation?
Our Response: ISA providers like Lambda School are already starting to negotiate directly with employers to ensure that students have a job after completing the curriculum. These relationships mitigate the risk of a student refusing to pay. Lambda School is able to do this because it’s developed such a strong curriculum. Furthermore, students face reputation risk should they try to avoid meeting their obligations to the ISA provider.
Future legislation should address instances where a student avoids payment or chooses to take a job with no salary (i.e., a student completes a coding bootcamp, but has a change of heart and goes to work at a non-profit that pays below the minimum income threshold.
Equity is expensive (relative to debt), so wouldn’t students be better off sticking with traditional debt financing?
Our Response: ISAs are not for everyone. ISA’s are best suited for people with greater expected volatility in their future earnings (instead of people with a strong likelihood of a certain amount of salary). This is similar to new businesses choosing between equity investment vs. debt to finance their operations. Businesses with clear expectations of future cashflows generally benefit more from debt vs. equity. Individuals looking to finance their education are no different. Similarly, ISA’s don’t need to be all or nothing. Individuals can choose to capitalize their education with a mix of student loans + ISA’s to get a more optimal mix.
Comments Off on Secondaries market growth underscores role of valuations – Robert Tribuiani
It’s not only limited partners scrutinizing valuations; regulators have also shown an interest amid the rise of the secondaries market and a growing propensity for GP-led transactions
By Robert A. Tribuiani, Managing Director, Head of Business Development, Murray Devine Valuation Advisors
For years, the secondaries market had been considered a small, sometimes-overlooked niche within private equity. After a year in which secondary investors amassed a record $74 billion of transaction volume, the momentum and growing impact of this “niche” can now be felt across the more expansive global PE landscape.
When the secondaries market first took root, the appeal was that it allowed investors to either sell limited partnership interests in existing funds or unload direct investments in companies within a captive portfolio. The upshot, of course, was that secondaries were able to offer liquidity in an otherwise illiquid asset class.
But as the secondaries market has grown in size, both limited partners and general partners are increasingly using the market to optimize their portfolio management. LPs, for instance, will use secondaries to continually refine their PE allocations, whether to minimize the J curve, improve diversification, or redeploy capital into newer funds with potential for more upside. GPs, alternatively, are turning to secondaries to effect restructurings in which existing LPs can exit a fund, knowing that new commitments bring additional capital and more time to fully realize remaining investments.
In a sense, secondaries have helped to further mainstream private equity and have brought a level of flexibility that appeals to both newer investors and most sophisticated LPs, alike. An unintended consequence, however, is that the growth of the secondaries market has also brought more scrutiny to valuations. This is particularly the case as GP-led transactions become more common.
As the proportion of GP-led secondaries have grown – by nearly a third year-over-year according to Greenhill & Co. – and the fact that they are becoming more common earlier in a funds life, the potential for conflicts of interest have investors and regulators scrutinizing valuations that much more closely. This should further dissuade GPs from attempting to smooth over any near-term volatility in their approach to valuation.
A recent Secondaries Investor article, for instance, highlighted some of the concerns raised by Oregon’s State Treasury in its annual private equity review. The pension noted that innovation in the secondaries market is translating into more complexity. Recaps involving more nascent partnerships can disrupt the alignment between GPs and LPs.
“While all of this has a place in a maturing private equity industry,” the pension noted, “the aggressive pace of innovation may suggest that secondary buyers have more appetite for deals than the current market can satisfy.” A corollary — one that brings to mind Ockham’s Razor – is that the innovation to accommodate this demand can also breed “complex and conflict-riddled transaction proposals,” the pension added.
That’s not to say that GP-led secondaries, on their own, are worthy of special attention or are in any way misleading. Still, many investors will use these transactions as an opportunity to revisit the mark-to-market valuations that usually serve as the basis for pricing secondary transactions.
By and large, though, the rise of the secondaries should be celebrated as an advance for the asset class. As the growth and increasing utility of the secondaries market continues in PE, so too does the transparency that aligns GP and LP interests. Ultimately, beyond providing confidence to investors and helping all constituencies manage risk, third-party independent valuations can also support ongoing growth of the secondaries market by imparting trust in the valuation that is helpful in navigating the added complexity.
Robert A. Tribuiani leads the overall business development efforts for Murray Devine Valuation Advisors. In this role, he is responsible for new business and works closely with the valuation client services team to support existing clients. Before joining Murray Devine, Rob worked as a senior business development executive for SolomonEdwards, a leading professional services firm headquartered in suburban Philadelphia and in similar roles for Longview Solutions and VerticalNet, both private equity backed companies. Rob is a graduate of Villanova University and earned a Bachelor of Science degree in Business Administration with a concentration in Finance.
Comments Off on Digital Transformation of Business and Society: Challenges and Opportunities by 2020 – Frank Diana
At a recent KPMG Robotic Innovations event, Futurist and friend Gerd Leonhard delivered a keynote titled “The Digital Transformation of Business and Society: Challenges and Opportunities by 2020”. I highly recommend viewing the Video of his presentation. As Gerd describes, he is a Futurist focused on foresight and observations — not predicting the future. We are at a point in history where every company needs a Gerd Leonhard. For many of the reasons presented in the video, future thinking is rapidly growing in importance. As Gerd so rightly points out, we are still vastly under-estimating the sheer velocity of change.
With regard to future thinking, Gerd used my future scenario slide to describe both the exponential and combinatorial nature of future scenarios — not only do we need to think exponentially, but we also need to think in a combinatorial manner. Gerd mentioned Tesla as a company that really knows how to do this.
He then described our current pivot point of exponential change: a point in history where humanity will change more in the next twenty years than in the previous 300. With that as a backdrop, he encouraged the audience to look five years into the future and spend 3 to 5% of their time focused on foresight. He quoted Peter Drucker (“In times of change the greatest danger is to act with yesterday’s logic”) and stated that leaders must shift from a focus on what is, to a focus on what could be. Gerd added that “wait and see” means “wait and die” (love that by the way). He urged leaders to focus on 2020 and build a plan to participate in that future, emphasizing the question is no longer what-if, but what-when. We are entering an era where the impossible is doable, and the headline for that era is: exponential, convergent, combinatorial, and inter-dependent — words that should be a key part of the leadership lexicon going forward. Here are some snapshots from his presentation:
Because of exponential progression, it is difficult to imagine the world in 5 years, and although the industrial era was impactful, it will not compare to what lies ahead. The danger of vastly under-estimating the sheer velocity of change is real. For example, in just three months, the projection for the number of autonomous vehicles sold in 2035 went from 100 million to 1.5 billion
Six years ago Gerd advised a German auto company about the driverless car and the implications of a sharing economy — and they laughed. Think of what’s happened in just six years — can’t imagine anyone is laughing now. He witnessed something similar as a veteran of the music business where he tried to guide the industry through digital disruption; an industry that shifted from selling $20 CDs to making a fraction of a penny per play. Gerd’s experience in the music business is a lesson we should learn from: you can’t stop people who see value from extracting that value. Protectionist behavior did not work, as the industry lost 71% of their revenue in 12 years. Streaming music will be huge, but the winners are not traditional players. The winners are Spotify, Apple, Facebook, Google, etc. This scenario likely plays out across every industry, as new businesses are emerging, but traditional companies are not running them. Gerd stressed that we can’t let this happen across these other industries
Anything that can be automated will be automated: truck drivers and pilots go away, as robots don’t need unions. There is just too much to be gained not to automate. For example, 60% of the cost in the system could be eliminated by interconnecting logistics, possibly realized via a Logistics Internet as described by economist Jeremy Rifkin. But the drive towards automation will have unintended consequences and some science fiction scenarios could play out. Humanity and technology are indeed intertwining, but technology does not have ethics. A self-driving car would need ethics, as we make difficult decisions while driving all the time. How does a car decide to hit a frog versus swerving and hitting a mother and her child? Speaking of science fiction scenarios, Gerd predicts that when these things come together, humans and machines will have converged:
Gerd has been using the term “Hellven” to represent the two paths technology can take. Is it 90% heaven and 10% hell (unintended consequences), or can this equation flip? He asks the question: Where are we trying to go with this? He used the real example of Drones used to benefit society (heaven), but people buying guns to shoot them down (hell). As we pursue exponential technologies, we must do it in a way that avoids negative consequences. Will we allow humanity to move down a path where by 2030, we will all be human-machine hybrids? Will hacking drive chaos, as hackers gain control of a vehicle? A recent Jeep recall of 1.4 million jeeps underscores the possibility. A world of super intelligence requires super humanity — technology does not have ethics, but society depends on it. Is this Ray Kurzweil vision what we want?
Is society truly ready for human-machine hybrids, or even advancements like the driverless car that may be closer to realization? Gerd used a very effective Video to make the point
Followers of my Blog know I’m a big believer in the coming shift to value ecosystems. Gerd described this as a move away from Egosystems, where large companies are running large things, to interdependent Ecosystems. I’ve talked about the blurring of industry boundaries and the movement towards ecosystems. We may ultimately move away from the industry construct and end up with a handful of ecosystems like: mobility, shelter, resources, wellness, growth, money, maker, and comfort
Our kids will live to 90 or 100 as the default. We are gaining 8 hours of longevity per day — one third of a year per year. Genetic engineering is likely to eradicate disease, impacting longevity and global population. DNA editing is becoming a real possibility in the next 10 years, and at least 50 Silicon Valley companies are focused on ending aging and eliminating death. One such company is Human Longevity Inc., which was co-founded by Peter Diamandis of Singularity University. Gerd used a quote from Peter to help the audience understand the motivation: “Today there are six to seven trillion dollars a year spent on healthcare, half of which goes to people over the age of 65. In addition, people over the age of 65 hold something on the order of $60 trillion in wealth. And the question is what would people pay for an extra 10, 20, 30, 40 years of healthy life. It’s a huge opportunity”
Gerd described the growing need to focus on the right side of our brain. He believes that algorithms can only go so far. Our right brain characteristics cannot be replicated by an algorithm, making a human-algorithm combination — or humarithm as Gerd calls it — a better path. The right brain characteristics that grow in importance and drive future hiring profiles are:
Google is on the way to becoming the global operating system — an Artificial Intelligence enterprise. In the future, you won’t search, because as a digital assistant, Google will already know what you want. Gerd quotes Ray Kurzweil in saying that by 2027, the capacity of one computer will equal that of the human brain — at which point we shift from an artificial narrow intelligence, to an artificial general intelligence. In thinking about AI, Gerd flips the paradigm to IA or intelligent Assistant. For example, Schwab already has an intelligent portfolio. He indicated that every bank is investing in intelligent portfolios that deal with simple investments that robots can handle. This leads to a 50% replacement of financial advisors by robots and AI
This intelligent assistant race has just begun, as Siri, Google Now, Facebook MoneyPenny, and Amazon Echo vie for intelligent assistant positioning. Intelligent assistants could eliminate the need for actual assistants in five years, and creep into countless scenarios over time. Police departments are already capable of determining who is likely to commit a crime in the next month, and there are examples of police taking preventative measures. Augmentation adds another dimension, as an officer wearing glasses can identify you upon seeing you and have your records displayed in front of them. There are over 100 companies focused on augmentation, and a number of intelligent assistant examples surrounding IBM Watson; the most discussed being the effectiveness of doctor assistance. An intelligent assistant is the likely first role in the autonomous vehicle transition, as cars step in to provide a number of valuable services without completely taking over. There are countless Examples emerging
Gerd took two polls during his keynote. Here is the first: how do you feel about the rise of intelligent digital assistants? Answers 1 and 2 below received the lion share of the votes
Collectively, automation, robotics, intelligent assistants, and artificial intelligence will reframe business, commerce, culture, and society. This is perhaps the key take away from a discussion like this. We are at an inflection point where reframing begins to drive real structural change. How many leaders are ready for true structural change?
Gerd likes to refer to the 7-ations: Digitization, De-Materialization, Automation, Virtualization, Optimization, Augmentation, and Robotization. Consequences of the exponential and combinatorial growth of these seven include dependency, job displacement, and abundance. Whereas these seven are tools for dramatic cost reduction, they also lead to abundance. Examples are everywhere, from the 16 million songs available through Spotify, to the 3D printed cars that require only 50 parts. As supply exceeds demand in category after category, we reach abundance. As Gerd put it, in five years’ time, genome sequencing will be cheaper than flushing the toilet and abundant energy will be available by 2035 (2015 will be the first year that a major oil company will leave the oil business to enter the abundance of the renewable business). Other things to consider regarding abundance:
Efficiency and business improvement is a path not a destination. Gerd estimates that total efficiency will be reached in 5 to 10 years, creating value through productivity gains along the way. However, after total efficiency is met, value comes from purpose. Purpose-driven companies have an aspirational purpose that aims to transform the planet; referred to as a massive transformative purpose in a recent book on exponential organizations. When you consider the value that the millennial generation places on purpose, it is clear that successful organizations must excel at both technology and humanity. If we allow technology to trump humanity, business would have no purpose
In the first phase, the value lies in the automation itself (productivity, cost savings). In the second phase, the value lies in those things that cannot be automated. Anything that is human about your company cannot be automated: purpose, design, and brand become more powerful. Companies must invent new things that are only possible because of automation
Technological unemployment is real this time — and exponential. Gerd talked to a recent study by the Economist that describes how robotics and artificial intelligence will increasingly be used in place of humans to perform repetitive tasks. On the other side of the spectrum is a demand for better customer service and greater skills in innovation driven by globalization and falling barriers to market entry. Therefore, creativity and social intelligence will become crucial differentiators for many businesses; jobs will increasingly demand skills in creative problem-solving and constructive interaction with others
Gerd described a basic income guarantee that may be necessary if some of these unemployment scenarios play out. Something like this is already on the ballot in Switzerland, and it is not the first time this has been talked about:
In the world of automation, experience becomes extremely valuable — and you can’t, nor should attempt to — automate experiences. We clearly see an intense focus on customer experience, and we had a great discussion on the topic on an August 26th Game Changers broadcast. Innovation is critical to both the service economy and experience economy. Gerd used a visual to describe the progression of economic value:
Source: B. Joseph Pine II and James Gilmore: The Experience Economy
Gerd used a second poll to sense how people would feel about humans becoming artificially intelligent. Here again, the audience leaned towards the first two possible answers:
Gerd then summarized the session as follows:
The future is exponential, combinatorial, and interdependent: the sooner we can adjust our thinking (lateral) the better we will be at designing our future.
My take: Gerd hits on a key point. Leaders must think differently. There is very little in a leader’s collective experience that can guide them through the type of change ahead — it requires us all to think differently
When looking at AI, consider trying IA first (intelligent assistance / augmentation).
My take: These considerations allow us to create the future in a way that avoids unintended consequences. Technology as a supplement, not a replacement
Efficiency and cost reduction based on automation, AI/IA and Robotization are good stories but not the final destination: we need to go beyond the 7-ations and inevitable abundance to create new value that cannot be easily automated.
My take: Future thinking is critical for us to be effective here. We have to have a sense as to where all of this is heading, if we are to effectively create new sources of value
We won’t just need better algorithms — we also need stronger humarithms i.e. values, ethics, standards, principles and social contracts.
My take: Gerd is an evangelist for creating our future in a way that avoids hellish outcomes — and kudos to him for being that voice
“The best way to predict the future is to create it” (Alan Kay).
My Take: our context when we think about the future puts it years away, and that is just not the case anymore. What we think will take ten years is likely to happen in two. We can’t create the future if we don’t focus on it through an exponential lens
Comments Off on Former Google CEO Eric Schmidt listed the ‘3 big failures’ he sees in tech startups today – Business Insider
Former Google CEO Eric Schmidt has listed the three “big failures” in tech entrepreneurship around the world.
Schmidt outlined the failings in a speech he gave at the Centre for Entrepreneurs in London this week. He later expanded on his thoughts in an interview with former BBC News boss James Harding.
Below are the three mistakes he outlined, with quotes taken from both a draft of his speech seen by Business Insider, and comments he delivered on the night.
1. People stick to who and what they know
“Far too often, we invest mostly in people we already know, who are working in very narrow disciplines,” Schmidt wrote in his draft.
In his speech, Schmidt pegged this point closely to a need for diversity and inclusion. He said companies need to be open to bringing in people from other countries and backgrounds.
He said entrepreneurship won’t flourish if people are “going to one institution, hiring only those people, and only — if I can be blunt — only white males.”
During the Q&A, Schmidt specifically addressed the gender imbalance in the tech industry. He said there’s a reason to be optimistic about women’s representation in tech improving, predicting that tech’s gender imbalance will vanish in one generation.
2. Too much focus on product and not on platforms
“We frequently don’t build the best technology platforms to tackle big social challenges, because often there is no immediate promise of commercial return,” Schmidt wrote in his draft.
“There are a million e-commerce apps but not enough speciality platforms for safely sharing and analyzing data on homelessness, climate change or refugees.”
Schmidt’s omitted this mention of socially conscious tech from his final speech, but did say that he sees a lot of innovation coming out of network platforms, which allow people to connect and pool data, because “the barrier to entry for these startups is very, very low.”
3. Companies aren’t partnering up early enough
Finally, Schmidt wrote in his draft that tech startups don’t partner enough with other companies in the modern, hyper-connected world. “It’s impossible to think about any major challenge for society in a silo,” he wrote.
He said in his speech that tech firms have to be ready to partner “fairly early.” He gave the example of a startup that wants to build homecare robots.
“The market for homecare robots is going to be very, very large. The problem is that you need visual systems, and machine learning systems, and listening systems, and motor systems, and so forth. You’re not going to be able to do it with three people,” he said.
After detailing his failures in tech entrepreneurship, Schmidt laid out what he views as the solution. He referred back to the Renaissance in Europe, saying people turned their hand to all sorts of disciplines, from science, to art, to business.
Comments Off on What Do Investors Need to Know About the Future of LED Grow Light Technology – Agfund
The horticultural lighting market is growing, and growing rapidly. According to a September press release from Report Linker, a market research firm specializing in agribusiness, the horticultural lighting market is estimated grow from a $2.43 billion market this year to $6.21 billion in 2023.
One of the key factors driving current market sector growth is increased development of LED grow light technology. LEDs (light emitting diodes) were first developed in the 1950s as a smaller and longer-lasting source of light compared to the traditional incandescent light bulb invented by Thomas Edison in 1879.
LEDs last longer, give off less heat, and are more efficient converting energy to light compared to other types of lights, all features that can result in higher yields and profits for indoor growers.
But until recently, LEDs were only used to grow plants indoors experimentally, largely because the cost was still too high for commercial businesses. Many commercial growers still use HID (High Intensity Discharge) lights such as High Pressure Sodium, Metal Halide, and Ceramic Metal Halide; all lights that have a high power output but are less durable than LED lights, generate far more heat, and have less customizable light spectra.
Today, LEDs are fast becoming the dominant horticultural lighting solution. This is due primarily to the one-million fold decrease in fabrication cost of semiconductor chips used to make LED lights since 1954.
For investors more familiar with field-based agriculture, it can certainly be a minefield to know where LED lighting technology for horticulture is going in the future. Although it is no longer the “early days” of LED technology development, current trends are still shaping the future of LED technology.
So what does the intelligent agtech investor need to know about the current state and future of LED grow light technology?
I interviewed Jeff Mastin, director of R&D at Total Grow LED Lighting, to discuss what the future of LED grow light technology for agriculture looks like, and how investors can use current trends to their advantage in the future.
What is your background – how did you get involved in grow light technology at Total Grow?
The company behind TotalGrow is called Venntis Technologies. Venntis has, and still does, specialize in integrating touch-sensing semiconductor technologies into applications.
Most people don’t realize LEDs are semiconductors; you can also use them for touch-sensing technologies, so there’s a strong bridge to agricultural LED technology.
Some of the biggest technical challenges in utilizing LEDs effectively for agriculture include LED glaring, shadowing and color separation.
We have used our expertise in touch-sensing LEDs to expand into horticultural LEDs, and we have developed technology that addresses the above challenges better, giving better control over the spectrum that the LED makes and the directional output of the light in a way that a standard LED by itself can’t do.
My personal background is in biology. When TotalGrow started exploring the horticultural world, that’s where being a biologist was a natural fit to take a lead on the science and the research side of the development process for the product; that was about 7 years ago now.
If you were going to distill your technical focus into trends that you’re seeing in the horticultural lighting space, what are the main trends to keep an eye on?
The horticultural lighting industry is really becoming revolutionized because of LEDs. Less than 10 years ago, LEDs in the horticultural world were mainly a research tool and a novelty.
In the past, they were not efficient enough and they were definitely not affordable enough yet to really consider them an economical general commercial light source.
But that is very quickly changing. The efficiencies are going up and prices down and they are really right now hitting the tipping point where for a lot of applications, but definitely not all applications, the LED world is starting to take over horticulture and indoor agriculture.
How do you view the translation of those trends into actionable points? For investors or technology developers in the agriculture technology space, how do they make sure that the LED light technology they are investing in isn’t going to be obsolete in a year or two?
With LEDs, the key question is still cost-efficiency, and there’s only so far the technology can improve.
Why? There are physical limitations. You can’t make a 100% efficient product that turns every bit of electricity into photons of light. At this point, the efficiency level of the top of line LEDs are up over 50%.
Can we ever get up to 70 or 80%? Probably not any time soon with an end-product, not one that’s going to be affordable and economical generally speaking.
So to answer your question, it’s not a category where you’re going to say, “well this is obsolete, I can get something three times better now.” The performance improvements will be more marginal in the future.
Ten years from now the cost will be cheaper. But that again doesn’t make current LED technologies obsolete. In terms of that fear, I don’t think people have to worry about current LED light technologies becoming obsolete.
In a large commercial vertical farming set up, what is the ballpark cost of horticultural LEDs currently?
To give just an order of magnitude sort of number, you’re probably going to be someplace in the $30 per square foot number for lights for a large facility. It can be half that or it can be double that.
That’s just talking within the realm of common vertical farming plants like greens and herbs, or other plants similar in size and lighting needs.
If you start talking about tomatoes or medicinal plants, then the ability to use higher light levels and have the plants make good use of it skyrockets. You can go four times higher with some of those other plants, and for good reason.
What type of horticultural lighting applications are LEDs still not the best solution for now and in the foreseeable future?
There are at least 3 areas where LEDs still may not make sense now and in the near future.
First, if the LED lights are not used often enough. The more hours per year the lights are used, the more quickly they return on their investment from power savings and reduced maintenance. Some applications only need a few weeks of lighting per year, which makes a cheaper solution appropriate.
Second, in some greenhouse applications, LED’s may not be the best choice for some time to come. Cheaper lights like high-pressure sodium have more of a role in greenhouses where hours of use are less and higher hang heights are possible. (Many greenhouses will still benefit strongly from LEDs, but the economics and other considerations make it important to consider both options in greenhouses.)
Lastly, some plants are not the best in vertical farming styles of growing where LEDs have their most drastic advantages. At least at this point it is not common to attempt to grow larger fruiting plants like tomatoes or cucumbers totally indoors, though when attempted that is still more practical with LEDs than legacy lights.
Comments Off on 6 Biases Holding You Back From Rational Thinking – Robert Greene
Emotions are continually affecting our thought processes and decisions, below the level of our awareness. And the most common emotion of them all is the desire for pleasure and the avoidance of pain. Our thoughts almost inevitably revolve around this desire; we simply recoil from entertaining ideas that are unpleasant or painful to us. We imagine we are looking for the truth, or being realistic, when in fact we are holding on to ideas that bring a release from tension and soothe our egos, make us feel superior. This pleasure principle in thinking is the source of all of our mental biases. If you believe that you are somehow immune to any of the following biases, it is simply an example of the pleasure principle in action. Instead, it is best to search and see how they continually operate inside of you, as well as learn how to identify such irrationality in others.
These biases, by distorting reality, lead to the mistakes and ineffective decisions that plague our lives. Being aware of them, we can begin to counterbalance their effects.
1) Confirmation Bias
I look at the evidence and arrive at my decisions through more or less rational processes.
To hold an idea and convince ourselves we arrived at it rationally, we go in search of evidence to support our view. What could be more objective or scientific? But because of the pleasure principle and its unconscious influence, we manage to find that evidence that confirms what we want to believe. This is known as confirmation bias.
We can see this at work in people’s plans, particularly those with high stakes. A plan is designed to lead to a positive, desired objective. If people considered the possible negative and positive consequences equally, they might find it hard to take any action. Inevitably they veer towards information that confirms the desired positive result, the rosy scenario, without realizing it. We also see this at work when people are supposedly asking for advice. This is the bane of most consultants. In the end, people want to hear their own ideas and preferences confirmed by an expert opinion. They will interpret what you say in light of what they want to hear; and if your advice runs counter to their desires, they will find some way to dismiss your opinion, your so-called expertise. The more powerful the person, the more they are subject to this form of the confirmation bias.
When investigating confirmation bias in the world take a look at theories that seem a little too good to be true. Statistics and studies are trotted out to prove them, which are not very difficult to find, once you are convinced of the rightness of your argument. On the Internet, it is easy to find studies that support both sides of an argument. In general, you should never accept the validity of people’s ideas because they have supplied “evidence.” Instead, examine the evidence yourself in the cold light of day, with as much skepticism as you can muster. Your first impulse should always be to find the evidence that disconfirms your most cherished beliefs and those of others. That is true science.
2) Conviction Bias
I believe in this idea so strongly. It must be true.
We hold on to an idea that is secretly pleasing to us, but deep inside we might have some doubts as to its truth and so we go an extra mile to convince ourselves — to believe in it with great vehemence, and to loudly contradict anyone who challenges us. How can our idea not be true if it brings out of us such energy to defend it, we tell ourselves? This bias is revealed even more clearly in our relationship to leaders — if they express an opinion with heated words and gestures, colorful metaphors and entertaining anecdotes, and a deep well of conviction, it must mean they have examined the idea carefully and therefore express it with such certainty. Those on the other hand who express nuances, whose tone is more hesitant, reveal weakness and self-doubt. They are probably lying, or so we think. This bias makes us prone to salesmen and demagogues who display conviction as a way to convince and deceive. They know that people are hungry for entertainment, so they cloak their half-truths with dramatic effects.
3) Appearance Bias
I understand the people I deal with; I see them just as they are.
We do not see people as they are, but as they appear to us. And these appearances are usually misleading. First, people have trained themselves in social situations to present the front that is appropriate and that will be judged positively. They seem to be in favor of the noblest causes, always presenting themselves as hardworking and conscientious. We take these masks for reality. Second, we are prone to fall for the halo effect — when we see certain negative or positive qualities in a person (social awkwardness, intelligence), other positive or negative qualities are implied that fit with this. People who are good looking generally seem more trustworthy, particularly politicians. If a person is successful, we imagine they are probably also ethical, conscientious and deserving of their good fortune. This obscures the fact that many people who get ahead have done so by doing less than moral actions, which they cleverly disguise from view.
4) The Group Bias
My ideas are my own. I do not listen to the group. I am not a conformist.
We are social animals by nature. The feeling of isolation, of difference from the group, is depressing and terrifying. We experience tremendous relief to find others who think the same way as we do. In fact, we are motivated to take up ideas and opinions because they bring us this relief. We are unaware of this pull and so imagine we have come to certain ideas completely on our own. Look at people that support one party or the other, one ideology — a noticeable orthodoxy or correctness prevails, without anyone saying anything or applying overt pressure. If someone is on the right or the left, their opinions will almost always follow the same direction on dozens of issues, as if by magic, and yet few would ever admit this influence on their thought patterns.
5) The Blame Bias
I learn from my experience and mistakes.
Mistakes and failures elicit the need to explain. We want to learn the lesson and not repeat the experience. But in truth, we do not like to look too closely at what we did; our introspection is limited. Our natural response is to blame others, circumstances, or a momentary lapse of judgment. The reason for this bias is that it is often too painful to look at our mistakes. It calls into question our feelings of superiority. It pokes at our ego. We go through the motions, pretending to reflect on what we did. But with the passage of time, the pleasure principle rises and we forget what small part in the mistake we ascribed to ourselves. Desire and emotion will blind us yet again, and we will repeat exactly the same mistake and go through the same mild recriminating process, followed by forgetfulness, until we die. If people truly learned from their experience, we would find few mistakes in the world, and career paths that ascend ever upward.
6) Superiority Bias
I’m different. I’m more rational than others, more ethical as well.
Few would say this to people in conversation. It sounds arrogant. But in numerous opinion polls and studies, when asked to compare themselves to others, people generally express a variation of this. It’s the equivalent of an optical illusion — we cannot seem to see our faults and irrationalities, only those of others. So, for instance, we’ll easily believe that those in the other political party do not come to their opinions based on rational principles, but those on our side have done so. On the ethical front, few will ever admit that they have resorted to deception or manipulation in their work, or have been clever and strategic in their career advancement. Everything they’ve got, or so they think, comes from natural talent and hard work. But with other people, we are quick to ascribe to them all kinds of Machiavellian tactics. This allows us to justify whatever we do, no matter the results.
We feel a tremendous pull to imagine ourselves as rational, decent, and ethical. These are qualities highly promoted in the culture. To show signs otherwise is to risk great disapproval. If all of this were true — if people were rational and morally superior — the world would be suffused with goodness and peace. We know, however, the reality, and so some people, perhaps all of us, are merely deceiving ourselves. Rationality and ethical qualities must be achieved through awareness and effort. They do not come naturally. They come through a maturation process.
Comments Off on Why Olam is Deploying Tech First, Then Thinking About CVC – AgFunder
Why Olam is Deploying Tech First, Then Thinking About CVC
“We have realized that some companies have gone down the wrong path by adopting the approach of inventing the problem. They find a technology that’s exciting and try to force-fit that technology for a problem that they don’t have. This is why we want to be very deliberate about the problems first, and then come to technology.”
“I’ll give you an example of blockchain. There’s so much hype about blockchain around the world. And in our industry, there are a few companies that have done some pilots. But we have not gone down that route, because we have not seen a tangible, scalable use case that could give us significant benefits for adopting blockchain.”
If one company could benefit from the efficiencies new technology can bring, it’s Olam, with a complex supply chain that grows, sources, processes, manufactures, transports, trades and markets 47 different agrifood products across 70 countries. These include commodities like coffee, cotton, cocoa, and palm oil that are farmed by over 4 million farmers globally, most of which are smallholders in developing countries.
But the third largest agribusiness in the world has been noticeably absent from the agrifood corporate venture capital scene in recent years, instead opting mostly to build its own technology solutions in-house. (It did deploy Phytech’s FitBit for crops in Australia in 2016 as an outside example.)
For traceability, and perhaps an alternative to blockchain-enabled technology, there’s Olam AtSource, with a digital dashboard that provides Olam customers with access to rich data, advanced foot-printing, and granular traceability. Olam hopes AtSource will help its customers “meet multiple social and environmental targets thereby increasing resilience in supply chains.”
Olam has also developed and deployed the Olam Farmer Information System (OFIS), a smallholder farm data collection platform providing smallholders with management tools and Olam customers with information about the provenance of products.
“OFIS solves the information issue by providing a revolutionary tech innovation for collecting and analyzing first mile data,” Brayn-Smith told AgFunderNews when OFIS launched in 2017. “We are able to register thousands of smallholders, GPS map their farms and local infrastructure, collect all types of farm gate level data such as the age of trees, and record every training intervention.”
This product is a clear example of a “transformational technology” that solves a problem for Olam and also gives the business efficiencies that could impact the bottom line, according to Sundararajan.
And Olam has built on top of OFIS to transact directly with cocoa farmers in Indonesia where Olam is publishing prices to around 30,000 farmers and buying cocoa directly from them.
“Before technology was available, it was almost impossible for any company to buy directly from the farmers, just because of the sheer volume and number of farmers. But, with technology, you have a far better reach, which will allow us to directly communicate with them,” Sundararajan tells AgFunderNews.
“Now the farmer can just accept a price and type in that he wants to supply it, and we arrange the complete logistics to pick up the cocoa from the farmer,” he says adding that the company’s country heads in other parts of the world are keen to launch this service in their markets. The company is starting next in Peru, then Guatemala, Colombia, Cote d’Ivoire, Ghana, and Nigeria.
Olam as Disruptor
While Olam deployed OFIS to solve for a problem, it also gives the company the opportunity to be disruptive in the markets it serves, according to Sundararajan.
As well as looking for transformational ways to solve specific problems, Olam also looks at “any ideas we have that will give Olam an opportunity to disrupt our own industry. So, we end up being a disrupter and not be at the risk of being disrupted by a new player,” he says.
“This fundamental shift in terms of Olam getting an opportunity to directly interact and transact with farmers is a starting point of disruption for us. This is a very complex point, which will bring into play several technologies for us to be able to successfully scale it.”
Going down this route, Sundararajan says Olam could end up providing farmers with new services and creating “separate streams of revenue that has nothing to do with what we were doing five or 10 years back.”
In this vein, Olam is working on deploying a technology to detect moisture — and therefore quality — in its commodities. The company is also looking at financial tools for its farmers.
“Looking at our business model, we believe that we have a few very good opportunities at the first mile of the supply chain and the last mile of the supply chain to change the way we compete,” says Sundararajan. “We believe that since we have control of the supply chain end-to-end, we can use technology to differentiate our service to customers in a way that our competitors will find difficult to replicate.”
Informal Startup Interactions
Olam does interact with startups on a selective basis, and Sundararajan’s participation in Rethink’s Singapore conference, as well as a hackathon it took part in with Fujitsu in Australia last year, are two examples. Sundararajan said he is considering an idea like The Unilever Foundry, but the company has yet to create a formal process or framework for these interactions. And the same goes for corporate venture capital.
“We believe that our digital journey has to mature much more, where we should demonstrate success within, by implementing the solutions that we’re developing, before even considering investing in venture capital. We believe that we have a very good strategy and a suite of products, stretching across from farm to the factories, to digitize our operations, whether it is a digital buying model, or whether it is spot factories in terms of predictive maintenance or increasing yield or it’s drone imagery from our own plantations, and productivity apps for employees.”
Comments Off on Building safe artificial intelligence: specification, robustness, and assurance – DeepMind
Building a rocket is hard. Each component requires careful thought and rigorous testing, with safety and reliability at the core of the designs. Rocket scientists and engineers come together to design everything from the navigation course to control systems, engines and landing gear. Once all the pieces are assembled and the systems are tested, we can put astronauts on board with confidence that things will go well.
If artificial intelligence (AI) is a rocket, then we will all have tickets on board some day. And, as in rockets, safety is a crucial part of building AI systems. Guaranteeing safety requires carefully designing a system from the ground up to ensure the various components work together as intended, while developing all the instruments necessary to oversee the successful operation of the system after deployment.
At a high level, safety research at DeepMind focuses on designing systems that reliably function as intended while discovering and mitigating possible near-term and long-term risks. Technical AI safety is a relatively nascent but rapidly evolving field, with its contents ranging from high-level and theoretical to empirical and concrete. The goal of this blog is to contribute to the development of the field and encourage substantive engagement with the technical ideas discussed, and in doing so, advance our collective understanding of AI safety.
In this inaugural post, we discuss three areas of technical AI safety: specification, robustness, and assurance. Future posts will broadly fit within the framework outlined here. While our views will inevitably evolve over time, we feel these three areas cover a sufficiently wide spectrum to provide a useful categorisation for ongoing and future research.
Specification: define the purpose of the system
You may be familiar with the story of King Midas and the golden touch. In one rendition, the Greek god Dionysus promised Midas any reward he wished for, as a sign of gratitude for the king having gone out of his way to show hospitality and graciousness to a friend of Dionysus. In response, Midas asked that anything he touched be turned into gold. He was overjoyed with this new power: an oak twig, a stone, and roses in the garden all turned to gold at his touch. But he soon discovered the folly of his wish: even food and drink turned to gold in his hands. In some versions of the story, even his daughter fell victim to the blessing that turned out to be a curse.
This story illustrates the problem of specification: how do we state what we want? The challenge of specification is to ensure that an AI system is incentivised to act in accordance with the designer’s true wishes, rather than optimising for a poorly-specified goal or the wrong goal altogether. Formally, we distinguish between three types of specifications:
ideal specification (the “wishes”), corresponding to the hypothetical (but hard to articulate) description of an ideal AI system that is fully aligned to the desires of the human operator;
design specification (the “blueprint”), corresponding to the specification that we actually use to build the AI system, e.g. the reward function that a reinforcement learning system maximises;
and revealed specification (the “behaviour”), which is the specification that best describes what actually happens, e.g. the reward function we can reverse-engineer from observing the system’s behaviour using, say, inverse reinforcement learning. This is typically different from the one provided by the human operator because AI systems are not perfect optimisers or because of other unforeseen consequences of the design specification.
A specification problem arises when there is a mismatch between the ideal specification and the revealed specification, that is, when the AI system doesn’t do what we’d like it to do. Research into the specification problem of technical AI safety asks the question: how do we design more principled and general objective functions, and help agents figure out when goals are misspecified? Problems that create a mismatch between the ideal and design specifications are in the design subcategory above, while problems that create a mismatch between the design and revealed specifications are in the emergent subcategory.
For instance, in our AI Safety Gridworlds* paper, we gave agents a reward function to optimise, but then evaluated their actual behaviour on a “safety performance function” that was hidden from the agents. This setup models the distinction above: the safety performance function is the ideal specification, which was imperfectly articulated as a reward function (design specification), and then implemented by the agents producing a specification which is implicitly revealed through their resulting policy.
*N.B.: in our AI Safety Gridworlds paper, we provided a different definition of specification and robustness problems from the one presented in this post.
As another example, consider the boat-racing game CoastRunners analysed by our colleagues at OpenAI (see Figure above from “Faulty Reward Functions in the Wild”). For most of us, the game’s goal is to finish a lap quickly and ahead of other players — this is our ideal specification. However, translating this goal into a precise reward function is difficult, so instead, CoastRunners rewards players (design specification) for hitting targets laid out along the route. Training an agent to play the game via reinforcement learning leads to a surprising behaviour: the agent drives the boat in circles to capture re-populating targets while repeatedly crashing and catching fire rather than finishing the race. From this behaviour we infer (revealed specification) that something is wrong with the game’s balance between the short-circuit’s rewards and the full lap rewards. There are many more examples like this of AI systems finding loopholes in their objective specification.
Robustness: design the system to withstand perturbations
There is an inherent level of risk, unpredictability, and volatility in real-world settings where AI systems operate. AI systems must be robust to unforeseen events and adversarial attacks that can damage or manipulate such systems.Research on the robustness of AI systems focuses on ensuring that our agents stay within safe limits, regardless of the conditions encountered. This can be achieved by avoiding risks (prevention) or by self-stabilisation and graceful degradation (recovery). Safety problems resulting from distributional shift, adversarial inputs, and unsafe exploration can be classified as robustness problems.
To illustrate the challenge of addressing distributional shift, consider a household cleaning robot that typically cleans a petless home. The robot is then deployed to clean a pet-friendly office, and encounters a pet during its cleaning operation. The robot, never having seen a pet before, proceeds to wash the pets with soap, leading to undesirable outcomes (Amodei and Olah et al., 2016). This is an example of a robustness problem that can result when the data distribution encountered at test time shifts from the distribution encountered during training.
Adversarial inputs are a specific case of distributional shift where inputs to an AI system are designed to trick the system through the use of specially designed inputs.
Unsafe exploration can result from a system that seeks to maximise its performance and attain goals without having safety guarantees that will not be violated during exploration, as it learns and explores in its environment. An example would be the household cleaning robot putting a wet mop in an electrical outlet while learning optimal mopping strategies (García and Fernández, 2015; Amodei and Olah et al., 2016).
Assurance: monitor and control system activity
Although careful safety engineering can rule out many safety risks, it is difficult to get everything right from the start. Once AI systems are deployed, we need tools to continuously monitor and adjust them. Our last category, assurance, addresses these problems from two angles: monitoring and enforcing.
Monitoring comprises all the methods for inspecting systems in order to analyse and predict their behaviour, both via human inspection (of summary statistics) and automated inspection (to sweep through vast amounts of activity records). Enforcement, on the other hand, involves designing mechanisms for controlling and restricting the behaviour of systems. Problems such as interpretability and interruptibility fall under monitoring and enforcement respectively.
AI systems are unlike us, both in their embodiments and in their way of processing data. This creates problems of interpretability; well-designed measurement tools and protocols allow the assessment of the quality of the decisions made by an AI system (Doshi-Velez and Kim, 2017). For instance, a medical AI system would ideally issue a diagnosis together with an explanation of how it reached the conclusion, so that doctors can inspect the reasoning process before approval (De Fauw et al., 2018). Furthermore, to understand more complex AI systems we might even employ automated methods for constructing models of behaviour using Machine theory of mind (Rabinowitz et al., 2018).
Finally, we want to be able to turn off an AI system whenever necessary. This is the problem of interruptibility. Designing a reliable off-switch is very challenging: for instance, because a reward-maximising AI system typically has strong incentives to prevent this from happening (Hadfield-Menell et al., 2017); and because such interruptions, especially when they are frequent, end up changing the original task, leading the AI system to draw the wrong conclusions from experience (Orseau and Armstrong, 2016).
We are building the foundations of a technology which will be used for many important applications in the future. It is worth bearing in mind that design decisions which are not safety-critical at the time of deployment can still have a large impact when the technology becomes widely used. Although convenient at the time, once these design choices have been irreversibly integrated into important systems the tradeoffs look different, and we may find they cause problems that are hard to fix without a complete redesign.
Two examples from the development of programming include the null pointer — which Tony Hoare refers to as his ‘billion-dollar mistake’– and the gets() routine in C. If early programming languages had been designed with security in mind, progress might have been slower but computer security today would probably be in a much stronger position.
With careful thought and planning now, we can avoid building in analogous problems and vulnerabilities. We hope the categorisation outlined in this post will serve as a useful framework for methodically planning in this way. Our intention is to ensure that AI systems of the future are not just ‘hopefully safe’ but robustly, verifiably safe — because we built them that way!
We look forward to continuing to make exciting progress in these areas, in close collaboration with the broader AI research community, and we encourage individuals across disciplines to consider entering or contributing to the field of AI safety research
Comments Off on How 20 big-name US VC firms invest at Series A & B – Pitchbook
NEA is one of the most well-known investors around, and the firm also takes the crown as the most active VC investor in Series A and B rounds in the US so far in 2018. Andreessen Horowitz, Accel and plenty of the other usual early-stage suspects are on the list, too.
Also included is a pair of names that have been in the news this year for backing away from the traditional VC model: Social Capital and SV Angel. The two are on the list thanks to deals completed earlier in the year.
Just how much are these prolific investors betting on Series A and Series B rounds? And at what valuation? We’ve used data from the PitchBook Platform to highlight a collection of the top venture capital investors in the US (excluding accelerators) and provide information about the Series A and B rounds they’ve joined so far this year. Click on the graphic below to open a PDF.
Comments Off on Corporate Venture Investment Climbs Higher Throughout 2018 – Crunchbase
Many corporations are pinning their futures on their venture investment portfolios. If you can’t beat startups at the innovation game, go into business with them as financial partners.
Though many technology companies have robust venture investment initiatives—Alphabet’s venture funding universe and Intel Capital’s prolific approach to startup investment come to mind—other corporations are just now doubling down on venture investments.
And 2018 is on track to set a record for U.S. corporate involvement in venture deals. We come to this conclusion after analyzing corporate venture investment patterns of the top 100 publicly traded U.S.-based companies (as ranked by market capitalizations at time of writing). The chart below shows that investing activity, broken out by stage, for each year since 2007.
A few things stick out in this chart.
The number of rounds these big corporations invest in is on track to set a new record in 2018. Keep in mind that there’s a little over one full quarter left in the year. And although the holidays tend to bring a modest slowdown in venture activity over time, there’s probably sufficient momentum to break prior records.
The other thing to note is that our subset of corporate investors have, over time, made more investments in seed and early-stage companies. In 2018 to date, seed and early-stage rounds account for over sixty percent of corporate venture deal flow, which may creep up as more rounds get reported. (There’s a documented reporting lag in angel, seed, and Series A deals in particular.) This is in line with the past couple of years.
Finally, we can view this chart as a kind of microcosm for blue-chip corporate risk attitudes over the past decade. It’s possible to see the fear and uncertainty of the 2008 financial crisis causing a pullback in risk capital investment.
Even though the crisis started in 2008, the stock market didn’t bottom out until 2009. You can see that bottom reflected in the low point of corporate venture investment activity. The economic recovery that followed, bolstered by cheap interest rates, that ultimately yielded the slightly bloated and strung-out market for both public and private investors? We’re in the thick of it now.
Whereas most traditional venture firms are beholden to their limited partners, that investor base is often spread rather thinly between different pension funds, endowments, funds-of-funds, and high-net worth family offices. With rare exception, corporate venture firms have just one investor: the corporation itself.
More often than not, that results in corporate venture investments being directionally aligned with corporate strategy. But corporations also invest in startups for the same reason garden-variety venture capitalists and angels do: to own a piece of the future.
A Note On Data
Our goal here was to develop as full a picture as possible of a corporation’s investing activity, which isn’t as straightforward as it sounds.
We started with a somewhat constrained dataset: the top 100 U.S.-based publicly traded companies, ranked by market capitalization at time of writing. We then traversed through each corporation’s network of sub-organizations as represented in Crunchbase data. This allowed us to collect not just the direct investments made by a given corporation, but investments made by its in-house venture funds and other subsidiaries as well.
It’s a similar method to what we did when investigating Alphabet’s investing universe. Using Alphabet as an example, we were able to capture its direct investments, plus the investments associated with its sub-organizations, and their sub-organizations in turn. Except instead of doing that for just one company, we did it for a list of 100.
This is by no means a perfect approach. It’s possible that corporations have venture arms listed in Crunchbase, but for one reason or another the venture arm isn’t listed as a sub-organization of its corporate parent. Additionally, since most of the corporations on this list have a global presence despite being based in the United States, it’s likely that some of them make investments in foreign markets that don’t get reported.