It reminds of this quote from Roger Penrose's book, Fashion, Faith, and Fantasy in the New Physics of the Universe:
“My nervousness was perhaps at its greatest because the illustrative area that I had elected to discuss, namely string theory and some of its various descendants, had been developed to its heights in Princeton probably more than anywhere else in the world.”
“Moreover, that subject is a distinctly technical one, and I cannot claim competence over many of its important ingredients, my familiarity with these technicalities being somewhat limited, particularly in view of my status as an outsider.”
“Yet, if only the insiders are considered competent to make critical comments about the subject, then the criticisms are likely to be limited to relatively technical issues, some of the broader aspects of criticism being, no doubt, significantly neglected.”
The fact that Penrose felt nervous criticizing string theory has made me think less of string theory (or rather, the humans behind it) ever since.
That's from 2003, when the string theory theorists were riding high and attacking string theory was bad for a physicist's career. Now, "with string theorists now virtually unemployable unless they can figure out how to rebrand as machine learning experts...", the situation has reversed.
String theorists understand high-dimensional math, so maybe they can do something for machine learning theory. Probably not, but we can hope. It's frustrating how much of a black box machine learning systems are.
Well... Penrose got himself into serious trouble speaking on issues beyond his expertise. I respect that he is now being more careful. And it's entirely possible that he isn't up to date on their tech. Why would you doubt his own words?
Thanks for illustrating my point. Eviscerating someone is not how science should be conducted. On the contrary, it's how we got here: decades invested in an untestable, unworkable theory.
I think the analogy here is with Grok generating images of (real) people wearing bikini. It could always be done in Photoshop before (and with hand-made photo montages before that), but it's now accessible at scale to people with zero skill. That's when a quantitative change becomes qualitative.
When people say efficient here, they mean cost efficient, extracting as much work per dollar from Anthropic as possible. This is the opposite of Anthropic’s view of efficiency, which would be providing the minimal amount of service for the most amount of money.
Nice. This reminds me of Logitech who open sourced LMS (Logitech Media Server) when they discontinued their multiroom product (known as Squeezebox before they bought it).
Still a fantastic multi-room setup to this day... I run a server as well as a client from a Raspberry Pi.
It's beautiful, but... I'm confused by the role of the play button. Can somebody explain how it works? I click on it and seemingly random things happen.
(no English source, unfortunately, but the title translates as: "“Useless and stupid”: French generative AI Lucie, backed by the government, mocked for its numerous bugs")
reply