I am at the TTI/Vanguard Next conference (agenda), with a sophisticated audience of tech executives from around the world. Of the topics I covered, the Q&A interest focused on iterative algorithms that will create an AI that exceeds human intelligence, much like biological evolution. (video)
Here are some of the related bullets from my slides:
Reed’s Law applies to combinations of ideas as well as self-forming groups. It’s the combinatorial explosion in the mating pool of ideas that creates perpetually accelerating progress.
Evolutionary algorithms allow us to build complex systems that exceed human understanding (synthetic biology, AI, innovative organizations), but there are some limitations to this approach:
• Subsystem Inscrutability
– Black box defined by its interfaces
– No “reverse evolution” (You can’t run that algorithm backwards)
• No simple shortcuts across the iterations
– Simulation ~ Reality
– Beauty from irreducibility
• Locus of Learning is Process, not Product
• Robust, within co-evolutionary islands
“The greatest achievement of our technology may well be the creation of tools that allow us to go beyond engineering – that allow us to create more than we can understand.” — Danny Hillis
“We actually think quantum machine learning may provide the most creative
problem-solving process under the known laws of physics.” — Google Blog
AI implications:
• Cut & Paste Portability?
• Locus of learning: Process, not Product
– Would we bother to reverse engineer?
– No hard take off?
•Co-evolutionary islands
– accustomed environment (differential immunity)
• Path dependence
– algorithm survival
– AI = Alien Intelligence defined by sensory I/O
Accelerating Technological Change
– Interdisciplinary Renaissance
– IT innervates $T markets
– More Black Swans
– Perpetual driver of disruption
==> Virtuous cycle for entrepreneurs
==> a great time for the new
Comments from others that followed:
“The majority of financial reports are now compiled by machines, not people.”
“A lot of the great data scientists are born in Russia, and they have the attributes of creativity, tenacity and an ability to code.”
“When we asked 1000 people on Mechanical Turk to flip a coin, we got 65% heads, 28% tails, and 7% typos. Many of them clearly did not actually flip a coin.”
“Imagine the sociological impact of crowdsourcing – what if you could create IBM for an afternoon and then disperse it? We might get cyber-Taylorism if we don’t think about doing it right.”
“Competition will be critical to the wisdom of crowds.”
Combinatorial Creativity: “Combinatorial search spaces are vast and the fastest supercomputers can not penetrate too deeply into them. Nevertheless, they may be able to penetrate several levels deeper than any person can, and thereby find superb creative acts that mankind did not or could not think of.”
Pointer to CHM video on the history of AI.
Photos by Ed Jay
Leave a Reply