🔮🌞 When will AI exceed humans? Self-driving car jam; moderating Facebook; the disappearing computer; R2D2, helium, spiders++ #115
|Azeem Azhar||May 28, 2017|
When will computers disappear? When will AI exceed humans? When will AI gain self-awareness? How do we deal with bots... and Facebook? When will self-driving cars materialise? Is coal really dead? And do spiders really think with their webs?
So many great questions!
🚀 Know someone who would enjoy this? Share via *|SHARE:facebook,twitter, linkedin|*
👍 Supported by SVB, Silicon Valley Bank
DEPT OF THE NEAR FUTURE
🌟 Walt Mossberg: The disappearing computer. In his last weekly column (ever), Mossberg reflects on where computing is going.
🤖 There are bots. Look around. Renee DiResta on bots and their role in information pollution:
So we’re at a point in which our marketplace of ideas bears striking resemblance to the financial markets in the early days of [high frequency trading]: deliberate manipulation, unanticipated feedback loops, and malicious algorithms are poisoning the ecosystem, introducing fragility and destroying confidence. But unlike in finance, it’s no one’s job to be looking at this. It’s no one’s job to regulate this.
🔮 Might AI lead to self-awareness in machines? And if it did, might we need to give it rights, argue Christof Koch and Giulio Tononi.
Special-purpose machines built following some of the same design principles as the brain, containing what’s called neuromorphic hardware, could in principle be capable of substantial conscious experience.
💸 Wall Street expects CEOs to articulate a compelling, disruptive future or face the axe, argues Christopher Mims in the WSJ.
🎧 PODCAST I had an outrageously interesting chat with Danny Lange, the VP of Artificial Intelligence and Machine Learning at Unity. He previously built Amazon's machine learning platform. We talk about how software paradigms and management systems will need to change. Oh, and when software engineers might become obsolete.
DEPT OF ARTIFICIAL INTELLIGENCE
When will AI exceed human performance? Fascinating Arxiv preprint from Katja Grace and collaborators:
Researchers predict AI will outperform humans in many activities in the next ten years, such as translating languages (by 2024), writing high-school essays (by 2026), driving a truck (by 2027), working in retail (by 2031), writing a bestselling book (by 2049), and working as a surgeon (by 2053). Researchers believe there is a 50% chance of AI outperforming humans in all tasks in 45 years and of automating all human jobs in 120 years, with Asian respondents expecting these dates much sooner than North Americans.
Of course, the limitations of this is that it is a best efforts prediction by expert practitioners (who may be too optimistic or too pessimistic, see the piece of IEA PV predictions below).
Google's AlphaGo is now the best Go player in the world. The new AlphaGo uses Google's tensor processing units, which are far more power-efficient than the previous GPU rig. (I heard 90% more power efficient somewhere but would prefer a better source.) AlphaGo also paired up with human Go players in an interesting example of intelligence augmentation.
Demis Hassabis of DeepMind on what is next for AlphaGo and its team of researchers who will now "develop advanced general algorithms that could one day help scientists as they tackle some of our most complex problems."
Apple is building a "Neural Engine", a custom chip for AI in its phones. The optimisations should lead to better battery life.
Intel showed off 70-fold speed improvements on training convolutional neural nets (on CPUs) by making software optimisations in Tensorflow. Algorithmic optimisations can often yield much better gains than hardware ones.
🤔 Should we worry about superintelligence or more grounded AI risks?
US Treasury secretary, Steve Mnuchin, doesn't think we need to worry about AI "kind of like R2D2 in Star Wars" impacting jobs.
Finland's basic income experiment is already making people feel better.
DEPT OF RENEWABLES
😎 "Experts view progression linearly, while tech quietly doubles and doubles and doubles". How the IEA has consistently under-forecast progress of photovoltaics.
"Coal is dead," says Blackrock, a $5.4trn asset manager. (Might this create a disinvestment spiral in that industry, ultimately, raising its cost of capital and making it progressively less economic?)
DEPT OF AUTONOMOUS VEHICLES
🐴 The Economist reckons autonomous vehicles are in the slow lane and further away from adoption than many think. (Great piece, but doesn't put its money on dates at which key milestones might be reached.)
🍕 Morgan Stanley revealed 30 stocks that stand to benefit from the autonomous car revolution, including Domino's Pizza.
What self-driving cars see. An overview of the LIDAR battle.
Toyota and BigChainDB team up to launch a project which allows firms to share autonomous vehicle driving data.
DEPT OF SOCIAL NETWORKS
🤦 Facebook's guide to content moderation was leaked by The Guardian. Just demonstrates that Facebook is, well, a media platform not a common carrier. And that they are making some pretty strong judgments which have previously governed not by fiat of a single firm (which is essentially controlled by a single individual) but by law or social consensus.
Facebook is increasing its number of content moderators to 7,500 (from 4,500), suggesting it still isn't taking this task very seriously. Facebook pays its moderators about $600 for a 40-hour week (according to this excellent reportage by Olivia Solon). The total moderation bill will run to about $250m per annum or about 1% of Facebook's profit.
A firm as important as Facebook, which enjoys increasing gross margins in excess of 85%, has enough latitude to invest more heavily in addressing this issue. The investment should be accompanied with greater data transparency and a more open, collegial approach to evaluating and dealing with the problem.
BBC: Facebook's tentacles reach further than you think and "what is most striking is the sense of resignation, the impotence of regulation, the lack of options, the public apathy."
No zone of privacy. Facebook "won't say if it will use your brain activity for advertising."
Due to dominant digital platforms, "markets today are radically different than what we believe – we have the façade of competition." (Review of a debate at Booth School of Business.)
America's FCC may honour anti-net neutrality comments left by bots.
SHORT MORSELS TO APPEAR SMART AT DINNER PARTIES
Rural America is the new inner city. A detailed look at drug abuse, rising early mortality, and joblessness.
Excellent analysis of Google & Apple Maps over a year puts Google's way ahead.
Challenges to overcome to get microrobots into human bodies. VERY COOL
The first cancer drug that targets mutation instead of location approved by the FDA.
A combination of nose job craze and an aging population drives new development trends in China.
Collection of 2.2 million pictures from the first 100 years of photography.
🎈 We need to better manage our precious supply of Helium.
Did I give a shout out to CogX London 2017? I'll be attending the event which will explore the impact of AI across industry. I have a 25% discount for the _Exponential View _community. Just visit and use the code 0849ar25conf.
I have a very old friend looking for a role in the startup ecosystem. He left a high-performing banking role a year ago to move into tech. I am certain, given the scale of his intellect and work rate, that he'd excel in a BD role for an early stage CEO or as an investor in a VC fund or family office. Preferably based in New York or London. Message me (to this email) if you'd like an intro.
This week's issue is brought to you with support from our partner, SVB