Dept of the near future
💯 “We are here to create”. Wonderful interview with Kai-Fu Lee, pioneering AI researcher and head of one of China’s Top VC firms. He’s been frank about the risks: “We’re all going to face a very challenging next fifteen or twenty years, when half of the jobs are going to be replaced by machines. Humans have never seen this scale of massive job decimation” and dismissive of the notion of the singularity which isn’t based on any “engineering reality”.
🔥 Facebook's senior exec, Andrew Bosworth’s leaked 2016 memo is a fascinating read. “We connect people. Period. That’s why all the work we do in growth is justified…[m]aybe someone dies.” He doesn’t use the word “utilitarian” or “consequentialist” but he describes essentially that. A single objective (growth) to be optimised for, and to hell with the consequences. (See also this eye-opening tale of how scammers use the Facebook ad platform.)
🌪️ A case for broadening the study of machine behaviour. These AI systems “are a new class of agents that inhabit our world. We must use every tool at our disposal to understand and regulate their impact on the human race.” (See also this long read: Francois Chollet, a deep learning expert, on his fears that AI could be used as a tool to exploit & manipulate people. Arguably it already has been.)
🕊️ EV reader, Hetan Shah, how we can use our personal data for the common good.
Dept of artificial intelligence
🇫🇷 Emmanuel Macron announced his intention to make France an AI leader and avoid “dystopia”, supported by €1.5bn in investment. Macron demonstrates nuanced understanding of the opportunity - both technological and social - of artificial intelligence in this must-read interview.
This sort of thing can only help capitalise on the value of these technologies. I do believe that cultural and intellectual diversity (and France has those when arrayed with China and the US) can only help in the development of appropriate AI systems. The equivalent UK number is only about £75m (€85m), which is a pity considering the nation's intellectual heritage across both humanities and technology domains, and bottom-up appetite for the Internet over the past 30 years. More importantly, few world leaders have expressed such adroitness with this sea change than Macron has, Obama being the last one to opine with such sagacity.
The last technological paradigms (software based on the Internet) were driven predominantly from the Silicon Valley and the US, which while not a monoculture, was not as diverse as the complex fabric of society these technologies now mediate.
I’m not a huge fan of thinking about this as a “race”. It implies that there is a finish line. In absence of good explanation of what that finish line is, perhaps it suggests super-intelligence or the singularity.
Much of the work done in AI is of the intangible quality, and so will drive spillovers. The knowledge is published, networked and disseminated widely wherever it is researched. We could all benefit.
In addition, the race to an AI state is not merely about harnessing data and smart maths-types. If you believe the hypothesis, then investing in the thinking about what society & businesses should look like after “AI” is also important. AI is a technology that could drive returns to superstars and severe inequality - which could lead to severe political risk, a.k.a. “pitchforks”. (More a consideration in the US than Europe). AI is a technology that glues together and layers on top of existing IT. Chinese medium and large enterprises have woefully underinvested in enterprise IT over the past decades. Is their fabric as ripe for AI enhancement as more pro-IT Western firms?
It’s a complex picture. We need to look further than data and doctorates and into how societies and polities can ready themselves for a world enabled by these technologies.
The Economist: How non-tech firms are starting to deploy AI at scale. (Reasonable survey.)
Microsoft is reorganising around AI. (It’s a press release. It’s in EV because I think its existence captures the magnitude of the opportunity.)
How AI engineers are tackling bias in face-recognition systems.
GE Medical Systems is figuring out how to deliver machine intelligence across its platforms. This great discussion highlights another important theme, the dispersion of compute across the edge and cloud.
The specs of the new Nvidia DGX-2 are impressive and continues the faster-than-Moore’s-Law progress we have been seeing in GPU and post-GPU architectures. (It’s becoming increasingly clear that AI systems need massive compute and memory. The memory needs to be super close to the compute and architectures will need to reflect that. The DGX-2 sports a switch which allows GPUs to access the same shared memory. But this approach is also power-hungry, which I think creates opportunities for entrepreneurs approaching this with a fresh piece of silicon.)
Dept of paying-it-forward 🐰
We welcomed our 26,000th reader this week—thanks to you. EV relies on your recommendations, so please take a moment this Easter Sunday to spread exponential love. Forward to a friend, tweet, or share on Linkedin.
Also, we're actively looking for partners for Q2 and Q3—if you lead a firm and would like to share your work with our audience, and in return support our newsletter, please email Marija.
Dept of transport and stuff
🙈 MIT’s Moral Machines survey, which investigates attitudes to a variety of trolley problems, may be the largest ethics survey in the world. Early results suggest we generally value saving youth over the elderly, and there is less consensus about complex tradeoffs.
There is a land grab in the bike sharing market underway in the US. One result? Abandoned bikes everywhere.
Is Tesla’s over-reliance on automation and robots hurting its product quality (and ultimately final output)?
Britain’s National Grid is confident that it can ready its infrastructure even if electric vehicles are dominating the market as soon as 2030.
The ethical quandaries of self-driving vehicles are far more complex than trolley problems.
Astro Teller, chief of Google X, describes how Waymo tests its cars in “pathological situations” which occasionally involve people in Elmo suits.
Israeli computer vision pioneer, Amnon Shashua analyses the Uber crash & its self-driving system and reckons his (shipping) safety systems would have prevented the accident.
💦 The most stressful job on the road may be to supervise autonomous vehicles. It is hard to stay vigilant.
Photos of 300,000 mothballed VW diesels.
Short morsels to appear smart at dinner parties
Americans spend $38bn on self-storage every year (rather than throwing stuff away or buying less?)
“Working out just how the future will work still depends on this interplay between fact and fiction.“ How science fiction writers have thought about the future of energy. (See also David Gerrold’s prescient 1999 predictions for the phone.)
🛡️ EV reader Michael Sulmeyer's level-headed analysis of the US cyber-offense.
How have 20 years on Viagra changed the American culture?
😮 The largest organ we never knew we had might harbour answers to how cancer spreads, among others.
💌 It takes 200 hours to make a friend.
Dropbox hit NASDAQ last week. A good time to reminisce about the original pitch deck which won over Sequoia's investment in 2007.
I’m looking into the area of intelligence augmentation and the notion of human:machine partnerships or “centaurs”. I’d love recommendations to research (current or historical) in this field. And if you are working in the field of IA, generative design, autonomous systems or collaborative robots, I’d love to chat.
If you are new to Exponential View, a brief reminder. I’m actively investing in great founders working on cool AI startups. If that is you, or someone you know, please send me a deck. I’m reasonably stage agnostic, so anything at Seed, Series A or B is of interest.
I’m off grid much of next week. Will endeavour to magic up an issue of the newsletter!
Sign in or become a Exponential View member to join the conversation.