A few months ago, I got a strange email with an attached Word file containing a story I can't adequately describe. The email instructed me to submit the piece for publication under my own name. I obliged. Today, it appears in Nature under the title Book of Cron Job. Take a look.
When they found it, they were confused. It was unusual to say the least. A rare format, hard to read. But with perseverance they translated its contents, and what they read changed their view of artificial intelligence forever.
#scifi by @suchow https://t.co/OVXqABCnWM
When they found it, they were confused. It was unusual to say the least. A rare format, hard to read. But with perseverance they translated its contents, and what they read changed their view of artificial intelligence forever.
#scifi by @suchow https://t.co/OVXqABCnWM
@ylecun@Noahpinion The disagreement is simple to resolve: there is no consensus definition of "intelligence". Being capable of learning huge amounts of declarative knowledge is widely considered a form of intelligence. And so is sample efficiency, learning so much from so little.
@kevinroose Weak criticism. The abstract's core claim is that user interactions moderate success of LLM deployments as medical assistants. Why would frontier models nix that interaction effect?
@krichard121212 The word "similar" is doing quite a lot of work here: immediately seeing the essence of a problem and being able to map it onto something you know is a big part of intelligence (see Hofstadter & Sander's Surfaces and Essences).
Ever read a news article about an NSF-funded nutritional epidemiology project and wondered why the supplement certification organization was creating complicated conflicts of interest for third-party testing of products by funding nutrition research? Wrong NSF.
fin.
I've always wanted to exit the classroom at that point and never come back in (it's the last day of class, anyways), but my wife says that this whole performance thing is already weird enough and that I shouldn't push my luck. Maybe one day I'll have the guts.
fin.
I teach an undergraduate course on judgment and decision making, where we focus on mathematical models of cognition — some simple, some complex.
I like to end the class with a performance of Dan Dennett's "Where am I?" (
https://t.co/Z8VRlIoWjQ).
1/x
The play ends with a dramatic change in narrator (sort of, you'll just need to read it to see what I mean), who walks of the stage to go revisit a clinic that's been discussed throughout.
4/x
That there may be past goals or dreams we have given up on prematurely, opportunities that we may have written off as missed, but are still there for the taking.
Or, put differently, "The best time to plant a tree was 20 years ago. The second best time is now."
fin.
But there's another part of me that wants to be inspired by quotes like these and so forces myself to reread it charitably, perhaps through a mental rewrite that changes the "never" to a "not always". It's a part of me that allows myself to actually take in its message:
4/x
Do doors really never close? Is there no such thing as an opportunity cost? Is every decision really reversible? No. We all make decisions that irreversibly shut off certain paths that our lives might have taken.
3/x
I sometimes minorly regret that my training as a scientist seems to have put me in what you might call "default skeptic" mode: rather than finding quotes like these inspirational, I immediately try to find a counterexample that topples it.
2/x