Diego Villafuerte S.

[Michael Seibel @mwseibel]

My default feed on Twitter is filled with too much snark, fear porn, and negativity. Is that a reflection of me, all of you, or a business trying to make ever increasing amounts of revenue with this product? I would literally pay for a more positive experience. And I think it would increase my faith in humanity. How much would it cost in lost advertising revenue to give me a more balanced feed? Name the price and I’ll pay. Otherwise all this addictive yelling at one another is eventually going to drive all of us insane.

[Andrej Karpathy @karpathy]

ChatGPT "Advanced Data Analysis" (which doesn't really have anything to do with data specifically) is an awesome tool for creating diagrams. I could probably code these diagrams myself, but it's soo much better to just sit back, and iterate in English.

In this example, I was experimenting with a possible diagram to explain Supervised Finetuning in LLMs. The "document" at the origin (0,0) is the empty document, and eminating outwards are token streams. Highlighted in black are the high probability token streams of the base model. In red are the token streams corresponding to the conversational finetuning data. When we finetune, we are increasing the probabilities of the red paths and suppressing the black paths. I like this view because it emphasizes LLMs as "token simulators", with their own kind of statistical physics backed by datasets, bouncing around in the discrete token space.

The conversation where we built it in a few minutes:
https://t.co/BPYipeQWws
(Sadly I just remembered that ChatGPT sharing doesn't support images, but at least the text is there, of me iterating with the diagram in plain language, and needing to touch no code. Such a vibe of the future.)

I had a similar experience yesterday, was trying to create a plot that shows smoothing in n-gram language models. Again I could just have coded this manually, but this was 10X faster and so easy.

Conversation:
https://t.co/MTxD2YH6Kv

Posting because during these chats I was struck again by that feeling of what must be the future, where you just sit back and say stuff, and the computer is doing the hard work. And in some narrow pockets of tasks, you can already get that feeling today.

[Diego Villafuerte S. @d_villafuerte]

@cwolferesearch Do you know of any examples evaluating fine-tuning with a few hundred thousand tokens in a different language? (Been looking for a short-term solution to this problem)

[Robin Hanson @robinhanson]

The current fertility situation:
https://t.co/pvF6zZ86Ze

[Diego Villafuerte S. @d_villafuerte]

Imaginate vivir en suiza y que los sistemas no salgan de vacaciones

[Diego Villafuerte S. @d_villafuerte]

@jp_villafuerte Está en beat features en tu perfil

[Diego Villafuerte S. @d_villafuerte]

El plug-in de code interpreter de ChatGPT es una locura absoluta. En 3 minutos pude manipular y analisar la base de unidades económicas DENUE de la CDMX sin abrir excel, ni escribir una linea de código. Me generó gráficas, explicaciones y me dio la base modificada. Increible.

[Diego Villafuerte S. @d_villafuerte]

No hay edit, pero analizar* 🙃

[Diego Villafuerte S. @d_villafuerte]

Dipping my toes in the AI assistant world with a simple wrapper app to log the nutritional info of everything I eat. See if it works for you! Will break at the slightest provocation so be gentle! https://t.co/LxXVnP1Bvh

[Sam Altman @sama]

i failed pretty hard at my first startup--it sucked!--and am doing pretty well on my second.

the thing i wish someone told me during the first one is that no one else thinks about your failures as much as you do, and that as long as don't psych yourself out you can try again.

[Bill Gross @Bill_Gross]

This is unbelievable to watch..

[The Global Fund @GlobalFund]

Working together @GlobalFund partnership has saved 44 MILLION lives since 2002. But for the first time in Global Fund history, key programmatic results for HIV, TB and malaria declined. #COVID19 is derailing progress. We cannot let that happen. https://t.co/Ex6aEXxKxS

[Bob Wachter @Bob_Wachter]

Covid (@UCSF) Chronicles,Day 521
I can't resist one more thread on boosters. I see lots of debate on data: about antibody levels, infections vs severe infctns, etc. I don't see much about the big issue: the tension between 2 perspectives – that of individuals vs. society.(1/25🧵)

[Timothy Burke @bubbaprog]

Simone Biles, in extreme slow motion.

[Austen Allred @Austen]

A fascinating story about when Thomas Watson Sr. (IBM CEO) was rambling in a speech.

His wife Jeannette decided to pass him a note.

[Aurelio Asiain🗣️ @aasiain]

Hay quien se blanquea; hay quien se ñerea. En ambos casos se nota el maquillaje y sólo creen o fingen creer que no se nota quienes creen en la simulación. Qué asco.

[Chris Fralic @chrisfralic]

Profile in Courage

[Paul Graham @paulg]

Banning plastic straws achieves nothing. Or at least, nothing good. It does achieve a couple bad things though: it gives people a false sense that something is being done to address ocean plastics, and misleads people about the true sources of them.

[Tim Urban @waitbutwhy]

In your next life, you'll be born into one of four societies. Knowing your position in the society will be luck of the draw, which would you hope to be born into?

Say poverty line is $10 in each case.

[Brian Armstrong @brian_armstrong]

The wild part is that you never know if you’re crazy, or everyone else is. The key here is that if your gut tells you there is something interesting, go take some action toward it and see what happens. Only action will produce more information, not endlessly debating it.