Ehsan Valavi

Postdoctoral Associate @MITSloan

[Ehsan Valavi @EhsanValavi]

Powerful integration of search and LLM agents! For long, I have been advocating for this and didn’t understand why we don’t have it!

[Felipe MIllon @Felipe_Millon]

Today, we at OpenAI launched Deep Researcher and I wanted to share a deeply personal story about how amazing this tool is and how it will change the world. Trigger warning, related to cancer....1/9

[Eric Schmidt @ericschmidt]

David Deutsch: "We have a duty to be optimistic. Because the future is open, not predetermined and therefore cannot just be accepted: we are all responsible for what it holds. Thus it is our duty to fight for a better world."

[Ehsan Valavi @EhsanValavi]

After a month of waiting, I called @FedEx for answers, but all I got was a chatbot that made me feel mocked! "Is this call regarding a delivery? …. Sorry, you cannot be connected..." Really, FedEx? #CustomerServiceFail #FedExFrustration

[Ehsan Valavi @EhsanValavi]

This is it! Thanks @ShengwuLi

[Shengwu Li @ShengwuLi]

Speculative take: The econ standard for coauthorship tracks the 'what money can't buy' distinction.

Some tasks can be compensated solely in cash. (e.g. we pay for copy editing services, access to datasets, participation in experiments, some forms of data collection.)

[Ehsan Valavi @EhsanValavi]

@eigenknight E^x^2 -e^x+x is convex with min=0

[Yann LeCun @ylecun]

Haha!
It wasn't quite like that.
Many senior members of the CV community were actually very friendly and curious about ConvNets (if skeptical).
There was a true intellectual debate.
Everyone had to have a "theory" of how vision worked.
For me, it was end-to-end gradient-based learning.
For some others, it was latent-variable Bayesian generative models.
But when the empirical evidence became incontrovertible, people updated their beliefs.

[Yann LeCun @ylecun]

The UAE Minister of AI @OmarSAlolama points to a historical precedent of premature technology regulation motivated by fear: the ban of the printing press in 1515 by Sultan Selim I led to the decline of the Ottoman Empire.

“We overregulated a technology, which was the printing press. It was adopted everywhere on Earth. The Middle East banned it for 200 years. The calligraphers came to the sultan and said: ‘We’re going to lose our jobs, do something to protect us’—so, job loss protection, very similar to AI. The religious scholars said people are going to print fake versions of the Quran and corrupt society—misinformation, second reason. It was fear of the unknown that led to this fateful decision."

https://t.co/CJQd6rMz5D

[Ehsan Valavi @EhsanValavi]

@AlphaSignalAI @johnjhorton

[Yann LeCun @ylecun]

@Noahpinion No.
The research arm of Bell Labs was never about moonshots.
It was about hiring the best scientists into small departments (typically 5 to 15 people) and giving them resources and a *lot* of freedom to work on what *they* deemed most promising.
That's how you get breakthroughs.

[Ehsan Valavi @EhsanValavi]

@navidas @JetBlue https://t.co/ZavMW7AS8k

[Andreas Steno Larsen @AndreasSteno]

The busiest day ever for commercial aviation yesterday

Almost 135k flights

Guess the tourism Industry is not in recession yet 😅

[Yann LeCun @ylecun]

Old joke.

[Yann LeCun @ylecun]

We all agree that we need to arrive at a consensus on a number of questions.
I agree with @geoffreyhinton that LLM have *some* level of understanding and that it is misleading to say they are "just statistics."
However, their understanding of the world is very superficial, in large part because they are trained purely on text.
Systems that would learn how the world works from vision would have a much deeper understanding of reality.
Second, auto-regressive LLM have very limited reasoning and planning abilities.
I do not believe we can get anywhere close to human-level AI (even cat-level AI) without
(1) learning world models from sensory inputs like video,
(2) an architecture that can reason and plan (not just auto-regress).
Now, if we have architectures that can plan, they will be *objective driven*: their planning will work by optimizing a set of objectives at inference time (not just training time).
These objectives can include guardrails that will make those system safe and subservient *even* if they end up having much better world models that humans.
Then, the problem becomes to design (or train) good objectives functions that will guarantee safety and efficiency.
It's a hard engineering problem, but not as hard as some have made it to be.

[Yann LeCun @ylecun]

"The future cannot be predicted, but futures can be invented" -- Dennis Gabor 1963.
"The best way to predict the future is to invent it" -- Alan Kay 1971

If you can't predict whether a technology is going to be beneficial or not, build it so it is.

https://t.co/JokWP3yPOF

[Ehsan Valavi @EhsanValavi]

Mission accomplished! 😎

[Harvard Business School @HarvardHBS]

It's Commencement Day! 🎉 Today's festivities began with a procession to Harvard Yard, joining the other schools of @Harvard for the conferral of degrees.

We are so proud of you, #HBS2023! 🎓

[Ehsan Valavi @EhsanValavi]

@grushkay Paris Hilton 😉
(Meant it as a joke)

Maison Albar Hotels Le Pont-Neuf is near attractions! Hope you have fun! 🤩
Link:
https://t.co/OGgp16zB5r

[David Baranger @David_Baranger]

Andrew Gelman famously said that "you need 16 times the sample size to estimate an interaction than to estimate a main effect" https://t.co/NoRJ4xI5Tq. Where does 16x come from? If you're used to thinking about interactions as regressions, his process is confusing! A 🧵...

[Ehsan Valavi @EhsanValavi]

What a week! 😅

[Seyed Moosavi @smoosavid]

“…most good ideas are obvious in retrospect, and humans are very bad at realizing that they just learned something when presented with such an idea.  We all react as, ‘oh yeah, of course, that’s true,’ not realizing that it wasn’t true to us an hour ago.”

[Bill Hilf @billhilf]

Strongly agree with @garymarcus in this well written piece 'Artificial General Intelligence Is Not as Imminent as You Might Think' https://t.co/o9G82K72cW. Less PR, more basic research.