Paul Scheid

@thomasgauvin super cool!

Cloudflare Mesh is here. Ready to connect your devices, servers, and agents to a single private network 🔐 And with Workers VPC, your Workers, Agents, and Durable Objects running on Cloudflare can now reach your private MCPs, APIs, and databases directly Oh, and it's 50 nodes + 50 users free on every account

gumloop raised a $50m series b led by benchmark here's a video we had fun making about the journey back to work.

@dwarkesh_sp +1 for @KordingLab

Looking for a neuroscientist to interview on my podcast. Keen for someone who can draw ML analogies for how the brain works (what's the architecture & loss/reward function of different parts, why can we generalize so well, how important is the particular hardware, etc).

@akshaykagrawal Cool! Gonna try this today.

Word on the street is ...

@arpangup Does this system successfully identify when students are using @cluely ?

99% of Edtech software does nothing to ensure students aren't cheating, guessing or rushing. This is a *huge* oversight that has made the majority of edtech apps ineffective for your average student. (cheating is rampant at schools + universities) For our Texas Sports Academy schools, we're solving this problem by building our own in-house cheat detection system. When the students are doing their 2 hours of academics, we have a recording system that watches their screen/webcam and streams the footage to our backend. All this video data gets analyzed by our AI system and any instances of cheating (in in-house or external learning apps) are immediately flagged for our Texas Sports Academy coaches & admins.

@karpathy Really enjoyed this interview! Would be willing to share some of the papers you’re referring to when you say “I've seen a number of papers pop up recently that are imo barking up the right tree along the lines of what I called "system prompt learning"”? 🙏

My pleasure to come on Dwarkesh last week, I thought the questions and conversation were really good. I re-watched the pod just now too. First of all, yes I know, and I'm sorry that I speak so fast :). It's to my detriment because sometimes my speaking thread out-executes my thinking thread, so I think I botched a few explanations due to that, and sometimes I was also nervous that I'm going too much on a tangent or too deep into something relatively spurious. Anyway, a few notes/pointers: AGI timelines. My comments on AGI timelines looks to be the most trending part of the early response. This is the "decade of agents" is a reference to this earlier tweet x.com/karpathy/statu… Basically my AI timelines are about 5-10X pessimistic w.r.t. what you'll find in your neighborhood SF AI house party or on your twitter timeline, but still quite optimistic w.r.t. a rising tide of AI deniers and skeptics. The apparent conflict is not: imo we simultaneously 1) saw a huge amount of progress in recent years with LLMs while 2) there is still a lot of work remaining (grunt work, integration work, sensors and actuators to the physical world, societal work, safety and security work (jailbreaks, poisoning, etc.)) and also research to get done before we have an entity that you'd prefer to hire over a person for an arbitrary job in the world. I think that overall, 10 years should otherwise be a very bullish timeline for AGI, it's only in contrast to present hype that it doesn't feel that way. Animals vs Ghosts. My earlier writeup on Sutton's podcast x.com/karpathy/statu… . I am suspicious that there is a single simple algorithm you can let loose on the world and it learns everything from scratch. If someone builds such a thing, I will be wrong and it will be the most incredible breakthrough in AI. In my mind, animals are not an example of this at all - they are prepackaged with a ton of intelligence by evolution and the learning they do is quite minimal overall (example: Zebra at birth). Putting our engineering hats on, we're not going to redo evolution. But with LLMs we have stumbled by an alternative approach to "prepackage" a ton of intelligence in a neural network - not by evolution, but by predicting the next token over the internet. This approach leads to a different kind of entity in the intelligence space. Distinct from animals, more like ghosts or spirits. But we can (and should) make them more animal like over time and in some ways that's what a lot of frontier work is about. On RL. I've critiqued RL a few times already, e.g. x.com/karpathy/statu… . First, you're "sucking supervision through a straw", so I think the signal/flop is very bad. RL is also very noisy because a completion might have lots of errors that might get encourages (if you happen to stumble to the right answer), and conversely brilliant insight tokens that might get discouraged (if you happen to screw up later). Process supervision and LLM judges have issues too. I think we'll see alternative learning paradigms. I am long "agentic interaction" but short "reinforcement learning" x.com/karpathy/statu…. I've seen a number of papers pop up recently that are imo barking up the right tree along the lines of what I called "system prompt learning" x.com/karpathy/statu… , but I think there is also a gap between ideas on arxiv and actual, at scale implementation at an LLM frontier lab that works in a general way. I am overall quite optimistic that we'll see good progress on this dimension of remaining work quite soon, and e.g. I'd even say ChatGPT memory and so on are primordial deployed examples of new learning paradigms. Cognitive core. My earlier post on "cognitive core": x.com/karpathy/statu… , the idea of stripping down LLMs, of making it harder for them to memorize, or actively stripping away their memory, to make them better at generalization. Otherwise they lean too hard on what they've memorized. Humans can't memorize so easily, which now looks more like a feature than a bug by contrast. Maybe the inability to memorize is a kind of regularization. Also my post from a while back on how the trend in model size is "backwards" and why "the models have to first get larger before they can get smaller" x.com/karpathy/statu… Time travel to Yann LeCun 1989. This is the post that I did a very hasty/bad job of describing on the pod: x.com/karpathy/statu… . Basically - how much could you improve Yann LeCun's results with the knowledge of 33 years of algorithmic progress? How constrained were the results by each of algorithms, data, and compute? Case study there of. nanochat. My end-to-end implementation of the ChatGPT training/inference pipeline (the bare essentials) x.com/karpathy/statu… On LLM agents. My critique of the industry is more in overshooting the tooling w.r.t. present capability. I live in what I view as an intermediate world where I want to collaborate with LLMs and where our pros/cons are matched up. The industry lives in a future where fully autonomous entities collaborate in parallel to write all the code and humans are useless. For example, I don't want an Agent that goes off for 20 minutes and comes back with 1,000 lines of code. I certainly don't feel ready to supervise a team of 10 of them. I'd like to go in chunks that I can keep in my head, where an LLM explains the code that it is writing. I'd like it to prove to me that what it did is correct, I want it to pull the API docs and show me that it used things correctly. I want it to make fewer assumptions and ask/collaborate with me when not sure about something. I want to learn along the way and become better as a programmer, not just get served mountains of code that I'm told works. I just think the tools should be more realistic w.r.t. their capability and how they fit into the industry today, and I fear that if this isn't done well we might end up with mountains of slop accumulating across software, and an increase in vulnerabilities, security breaches and etc. x.com/karpathy/statu… Job automation. How the radiologists are doing great x.com/karpathy/statu… and what jobs are more susceptible to automation and why. Physics. Children should learn physics in early education not because they go on to do physics, but because it is the subject that best boots up a brain. Physicists are the intellectual embryonic stem cell x.com/karpathy/statu… I have a longer post that has been half-written in my drafts for ~year, which I hope to finish soon. Thanks again Dwarkesh for having me over!

Don’t worry, there is an AI lab actually working on cancer.

@levelsio @veggie_eric +1

@veggie_eric Most food in SF is really average I am not some pretenious foodie but anywhere else it's just better I think they just work too hard to care much about food taste

why is pizza in the Bay Area such absolute trash like wtf, we nerds can cook up a recipe for a gajillion-parameter model no problemo but can't figure out a mildly appetizing ratio for some flour, water, and cheese

@natolambert yeah, love Waymo but hard to beat the efficiency of taking BART to SFO, esp if you live near a station in sf. Plus, BART has one of these:

The queues for this are going to be insane. Would you wait 40min for a Waymo or just take BART?

@Gena_I_Gorlin neat! sounds similar in principle to simcare.ai. Curious if you’ve tried it, or competitors, and if so how the interactions compare (beyond the difference btw avatar vs. text).

Coolest AI use case I’ve unlocked to date: students practice therapy on a bot that plays a specific patient (based on fictionalized case report I’ve provided) They can “pause role play” anytime to ask research questions or get feedback I get the transcript to review and grade

@aidan_mclau i appreciate the use of the semicolon

@sashadoubov Ha, your tweet came right after this one on my tl:

the ai researcher’s urge to add pre to the thing you’re actually doing (pretraining, prefill)

@Marty_J_Moon @EMostaque You can also access ChatGPT via WhatsApp: chatgpt.com/link/whatsapp/.

@EMostaque I was thinking this morning that if I could iMessage or WhatsApp an AI assistant like my friends and family, it would be more fluid than having to go in and out of an app.

What is the absolute best chat ui/ux or experience for LLMs? Anyone cracked this?

I'm excited to share my work on the Virtual Lab, a collaboration between LLM-powered AI agents and human researchers to identify nanobody binders to SARS-CoV-2, published in @Nature. nature.com/articles/s4158… Our code is publicly available at github.com/zou-group/virt…

Delicious prompt sando from @alxai_ at @every’s Claude Code Camp⛺️. So many insights...excited for the next one!

🚀💫 We are so proud to have our @TwistBioscience PGx panel long-read @nanopore data to be used as part of the upcoming FREE Virtual Training Project offered by @MilrdOrg with @nvidia & @DeepOriginBio! We also have the great opportunity for some of our CariGenetics interns #MatthewWest, #MicaelaMcLean & #SandraWatt to be Assistant Mentors for this Webinar, helping those who participate. 👨🏾‍💻👩🏾‍💻👩🏼‍💻 By being apart of this VTP, we have ensured the inclusion of diverse data that represents Caribbean genomics 🏝️ Are you an aspiring bioinformatician or data scientist interest in applying your skills to genetic data? This Webinar is for just for you 🙌🏾 Sign-up here ✍️: deeporigin.com/register-for-o… Please RT to spread the word 🗣️

@Meliseymo @sarah_shapley @jarildy @rstats4ds @RLadiesGlobal @BlackGirlsCode @GirlsWhoCode @sarah_shapley thanks for the shoutout! @Meliseymo happy to chat about @MilrdOrg’s VTPs if you’re interested:)

@sarah_shapley @jarildy @rstats4ds @RLadiesGlobal @BlackGirlsCode @GirlsWhoCode @paulscheid Thank you so much for this!!!

What were some of the most helpful packages, programs or tips you learned when you first started doing bulk RNAseq and single cell seq analyses? @rstats4ds @RLadiesGlobal @BlackGirlsCode @GirlsWhoCode #BlackGirlsCode #RNAseq

I did the same VTP training with @JustinAPater from @MilrdOrg and found it incredibly useful to learn how to QC scRNA-seq data (&more) in a more realistic aws environment. Lots of hands-on help and a fantastic learning experience. Special thx to @paulscheid!

Thank you @JustinAPater! Science Twitter: we still have slots for this and other VTPs available. Unlimited support + your own @awscloud instance + help/advice with your own data Schedule + Enrollment Info here: milrd.org/schedule-enrol…