Discussion about this post

User's avatar
jabowery's avatar

The perennial misunderstanding I encounter as one of the Hutter Prize administrators is that entries to the Hutter Prize represent the application of "artificial intelligence" to generate incisive knowledge representation. Entries to the Hutter Prize are incisive knowledge representations -- purported to be more incisive, hence PREDICTIVE (by algorithmic information theory), than other entries. The proof consists merely of executing the entries, running a checksum on the output and comparing the sizes of the entries: No "AI" necessarily be involved. Humans may have written a program that, when executed, achieves the same thing, and submit it as an entry.

With very few exceptions, EVERYONE involved with AI, from Hinton to Agrawal, who has any visibility falls victim to this misunderstanding in one way or another even though it is an obvious consequence of the most widely-accepted theorem of artificial general intelligence -- algorithmic probability. The implication is that the people parading around as experts in AI don't understand the founding theory.

Expand full comment
efalken's avatar

Online behavior (Netflix, Facebook, Amazon) is amenable to very nonlinear algorithms optimized over millions of datapoints. To extrapolate this to what most people do, however, is silly. Stock analysis, accounting, construction, etc., has been and always will be working at the margin to improve productivity, and luckily, productivity does generally increase, which is a good thing.

Replacing humans with machines has been going on for over 200 years (eg, Luddites), and there have always been those who think soon people will be obsolete (see Keynes' Economic Possibilities for our Grandchildren, or Charlie Chaplin's Modern Times). They have always been wrong.

The main reason low-capital Americans have been losing out is twofold. First, there's more welfare, in terms of claiming Disability under Social Security, and the fact that basic necessities like food and clothing are so cheap, so more can afford to never enter the labor market, neglecting the long-term affect of this on their human capital (building skills, including discipline). Second, low-skilled immigration has more than offset this, keeping low skilled wages low (immigration fell from 1910-1970, increased since).

Expand full comment

No posts