š®Ā AI is just beginning; global governance; the returns on deep learning; AI pay goes exponential, dolphin people & ketamine++ #162
š Summer is here! And youāve got a bumper issue of Exponential View. Find a hammock in some shade, and enjoy it! Get all the way to the end, you wonāt be disappointed.
Dept of the near future
š®Ā The AI revolution hasnāt happened, yet. Long, must-read: we need a 'human-centric engineering discipline'Ā to ensure that what we develop helps humans flourish.
šøĀ Ā Advanced AI techniques (such as deep learning and reinforcement learning) could unlock around $5 trillion of value annuallyĀ across multiple industries according to the McKinsey Global Institute. (See also, demographics matter: in low-population growth nations, like those of Eastern Europe,Ā robots are filling gaps in employment.)
Ā
šĀ Steve Sinofsky: New breakthroughs canāt do new things in old ways. What incumbents and commentators get wrong about step-change innovations.
š„Ā We must regulate modern cyber warfare before these e-skirmishesĀ turn into hot wars (and before our AI systems get more capable), argue Luciano Floridi & Mariarosaria Taddeo.
šĀ Bill Gates: CRISPR could help humanity overcome some of the biggest and most persistent challenges in global health & development. (See also: Goldman Sachs notes that gene therapies that cure patients could challenge the business models of firms dependent on recurrent sales of drugs.)
Dept of states and sovreignty
There is a notion worth revisiting: are nation states nearing their end as our preferred scale of theĀ political and socio-economic organisation? This idea lies in contrast with the āend of historyā theory of modernity.
Iāve found the question of evolving the nation-stateĀ fascinating since the early 1990s when I first came across ideas of decentralised forms of organisation enabled by new electronic networks (and was contemporaneously studying political institutions and models of governance).
Of course, recently we've seen resurgent nationalism in Turkey, the US, Britain, Hungary and many other places.Ā This might challenge the 'end of the nation-state' thesis, this nationalism isĀ a sort ofĀ reversion to the mean. But could it, instead, be the febrileĀ twitching that presages rigor mortis?
Rana Dasgupta argues this compellingly:
20th-century political structures are drowning in a 21st-century ocean of deregulated finance, autonomous technology, religious militancy and great-power rivalry. Meanwhile, the suppressed consequences of 20th-century recklessness in the once-colonised world are erupting, cracking nations into fragments and forcing populations into post-national solidarities
As Iāve argued in Exponential View and elsewhere, we increasingly need to adapt our existing institutions or invent new ones in order to cope with the changes in our economies, demographics, natural resources and climate.
As Dasgputa concludes: āThis is not a small endeavour: it will take the better part of this century. We do not know yet where it will lead.ā
One example is the relationship between the corporation and the states, and the balance between their power. Almost a decade ago my friend and researcher, Stephanie Hare, introduced me to the idea of ācorporate foreign policyā, the notion that states needed to formally recognise that the emerging dominance of technology platforms merited a formal quasi-diplomatic status. Some years later, the forward-thinking Danes appointed Casper Klynge as the worldās first Technology Ambassador, in the vein first described to me by Stephanie several years earlier.
In other cases, governments seem unwilling or incapable to execute on their duties - for example, maintaining their integrity in the face of or responses to cyber attacks (which are not really any different from other types of hostility). In those situations, private corporations are stepping in to fill or exploit the vacuum.
Some interesting points have emerged in the past few weeks:
Thirty-four tech companies accord to protect end users from cyber attacks from any source and not to support national actors in cyber offense.
Microsoft disclosed it has refused to sell AI tools to customers it believed had bad intentions. (The boundaries of what one can and cannot sell are normally prescribed by the law. Companies have often strictly adhered to Milton Friedmanās 1970 expectoration that a firm had āonly one social responsibility ...to use it[s] resources and engage in activities designed to increase its profits so long as it stays within the rules of the game.ā)
Bloomberg explores the scale of Palantirās knowledge of the average American in an eye-opening feature: āWhen whole communities are algorithmically scraped for pre-crime suspects, data is destiny.ā
Profile of ML professor Pedro Domingos onĀ national competition in artificial intelligence, the advance of autocrats and the threats modern technology presents to Western democracies.
Dept of artificial intelligence
Iāve written previously about how Mooreās Law is petering out, and argued that the demands of AI will provide the correct incentives to drive chip performance. Now some analysts have coined Huangās Law (after Jensen Huang, founder of Nvidia) articulating this one mechanism to explain why we continue to see faster-than-Mooreās Law progress in processor power/performance.
FacebookĀ is developing its own chips. They could use the chips to power hardware devices, artificial intelligence software and servers in its data centers. (The company joins Google, Apple, Amazon and Microsoft in now developing their own chips optimised for AI applications, alongside more than a dozen new entrants.)
Several EV readers were involved in the UKās House of Lords AI report. Available in depth here. A good summary is here.
āThe IEEE wants to understand the perspectives of female A/IS practitioners from a cultural and technological standpoint. Participate in the survey HERE.
šÆĀ We are already all cyborgs: Fascinating profile of philosopher Andy Clark, a key proponent of the extended mind hypothesis.
Nearly 60% of corporate execs have noĀ dedicated investments or plansĀ to retrain their staff as legacy jobs are eroded by automation.
The FDA approved first AI diagnostic device that doesnāt need a specialist to interpret the results.
š¤Ā After few failed attempts robots put together an Ikea chair in 8 minutes.
Short morsels to appear smart at dinner parties
Top AI researchers are being paid more than $1m a year. (Elsewhere, the median annual pay at McDonaldās is $7k, at Facebook $240k.)
Brilliant physicist, Carlo Rovelli, on the meaning of time. (FT paywall)
How decentralised are crypto networks? (Remarkable data set.)
Bitcoin declared compliant with Sharia law.
šĀ Ā Bajau sea nomads (humans)Ā have evolved for deeper, longer diving.
Chief execs with broad, wide networks create more firm value.
Russian had to block 16m IP addresses to ban Telegram.
š·Ā Ā The average cost of stem-cell treatment isĀ between $7,500-$10,000.
ā»ļøĀ ScientistsĀ accidentally created an enzyme that eats PET.
Hackers stole casinoās database through aĀ fish tankĀ thermometer.
š Britain went 55 hours without burning any coal. A new record.
End note
As I may have mentioned previously, Iām looking for breakthrough thinkers in the AI domain for The Cutting Edge, the future stream of Londonās CogXĀ AI Festival on June 11-12. Iām particularly interested in finding anyone building novel chips or tools for AI systems; new research frontiers beyond deep learning (e.g. evolutionary approaches, reinforcement learning, symbolic approaches, reasoning etc.) as well as breakthrough applications of the technology (for examples, in areas like disease detection, materials science, manufacturing, decision making, simulation & more.)
The schedule is pretty packed, but if you know a researcher, engineer, analyst or storyteller, doing amazing work in this area, please let me know.Ā
Cheers,
Azeem
P.S. As the summer is approaching, recommend Exponential View to your friends who are planning active vacationsāit's an enriching read for the warmest months. Twitter counts, too.