François Chollet on X: "I would have assumed it was fairly obvious, but in case it's not: a million-line codebase (also known as a "harness"), running at inference time, orchestrating thousands of calls to a neural network for any given task, is the exact definition of a "neurosymbolic architecture"" / X<br>Post
Log inSign up
Post
François Chollet
@fchollet
I would have assumed it was fairly obvious, but in case it's not: a million-line codebase (also known as a "harness"), running at inference time, orchestrating thousands of calls to a neural network for any given task, is the exact definition of a "neurosymbolic architecture"<br>span:not(:empty)~span:not(:empty)]:before:content-['·'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0">11:13 AM · Aug 6, 202698.5KViews
82<br>112<br>1.5K<br>548
span:not(:empty)~span:not(:empty)]:before:content-['·'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">François Chollet
@fchollet
4h
Among proponents of neurosymbolic architectures, there had been some debate over the years about whether the outer level would be symbolic (i.e. a harness that calls neural models) or whether the outer level would be neural (i.e. a neural model using symbolic tool calls as part Show more
12<br>17<br>252<br>21K
span:not(:empty)~span:not(:empty)]:before:content-['·'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">François Chollet
@fchollet
3h
For a very long time most high-performing AI models were end-to-end neural models; vector input -> vector output, with only ultra-thin symbolic preprocessing and postprocessing layers (e.g. label decoding). For many, it seemed that moving more and more logic to the end-to-end Show more
16<br>10<br>194<br>26K
span:not(:empty)~span:not(:empty)]:before:content-['·'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">François Chollet
@fchollet
35m
Some folks already saying "no no it's only neurosymbolic if the symbolic part encodes cognitive logic, otherwise that's just engineering" my man what do you think is in the million-loc harnesses. Why do you think they're this large. Why do you think their presence is critical for Show more
21<br>4K
span:not(:empty)~span:not(:empty)]:before:content-['·'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">François Chollet
@fchollet
15m
An accurate characterization of the arc of AI is that it is shaped by two trends:
1. Moving more and more logic to a neural model for tasks where training data can be densely sampled (e.g. the shift from pre-DL feature engineering to end-to-end learning circa 2013-2016, and more Show more
22<br>2.5K
span:not(:empty)~span:not(:empty)]:before:content-['·'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">François Chollet
@fchollet
1m
A third trend that you should expect in the future is that the symbolic layers will be learned / evolved, rather than hand-engineered.
319
span:not(:empty)~span:not(:empty)]:before:content-['·'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">🍓🍓🍓
@iruletheworldmo
3h
i don’t get why you want to test an outdated llm that no one will ever use again.
you should be making benchmarks that can’t be saturated by anything.
what’s interesting about benchmarking a severely hamstrung model?
31<br>3K
Log in or sign up for X<br>See what’s happening and join the conversation<br>Continue with phoneContinue with AppleContinue with Google<br>or<br>Log in with username or email
Relevant people
François Chollet@fcholletFollow<br>Co-founder @ndea. Co-founder @arcprize. Creator of Keras and ARC-AGI. Author of 'Deep Learning with Python'.
Trending now