Loading activity…
@chrszegedy
#deeplearning, #ai research scientist. Opinions are mine.
AI
A user frames the choice around how an AI’s intelligence compares with their own.
AI
A post argues that training AI for elegant proofs would be far easier than teaching it to produce valid ones.
AI
A post describing a Meta paper says the model uses a network of associative memories instead of standard attention.
AI
A post argues most AI progress measures tracked exponential growth in 2023–2026, while capital spending grew slightly super-exponentially.
AI
A poster says they expect economic effects over the next two years if AI is paused for 10 years.
AI
A user shared the statement; another criticized people they said were attacking it while defending OpenAI.
AI
A user wonders what Gary Marcus thinks of the bet now and says they saved screenshots from a seemingly deleted thread.
AI
AcerFur used the line while quoting an earlier post, without explaining what “it” refers to.
AI
A researcher recalls three attendees claiming a speaker’s work had been done a few, 10 and 20 years earlier.
AI
OpenAI says it drew on an independent mathematics advisory group’s advice about how to release the results.
AI
In an exchange on X, tszzl argued mechanistic interpretability is a minimum requirement for turning AI alignment into an engineering discipline. Neel Nanda joked about the tradeoff, and Geoffrey Irving pushed back on calling any current subfield necessary.
AI
Elon Musk used the phrase ‘biological bootloader’; another poster recalled him linking a similar view from Larry Page to starting OpenAI.
AI
A post quotes AI as a 'superhuman specialist' that's 'sub-human at running itself,' then speculates about what that gap could lead to.
AI
A user calls the predicted wait the “Navier-Stokes effect,” saying Nobel committees won’t consider them for prizes anytime soon.
AI
A user asked the AIs to budget for a family of four starting on $150,000 gross income and calculate inflation from 2006 to 2026.
AI
A user also claims Astra solved five Erdős problems, two under budget.
AI
A user endorses Kevin Buzzard’s proposal to let machines explore math, then begin the journey toward wherever they get stuck.
AI
The author says one of their main recommendations is to post results as quickly as possible—meaning once certain minimum standards have been met.
AI
One post argues that no technology in history has accumulated more “hacks” faster than AI.
AI
One participant argues that generative architectures, especially those producing discrete symbols, are ill-suited to understanding reality. A reply offers floating-point numbers and transformer-based video generators as counterarguments.
