Maja Augustynek

Order and chaos provider

[Maja Augustynek @MajaAugustynek]

I think they should finally move to EU…

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[Maja Augustynek @MajaAugustynek]

@AnthropicAI https://t.co/5KOIxkoSdF

[Andrej Karpathy @karpathy]

I'm being accused of overhyping the [site everyone heard too much about today already]. People's reactions varied very widely, from "how is this interesting at all" all the way to "it's so over".

To add a few words beyond just memes in jest - obviously when you take a look at the activity, it's a lot of garbage - spams, scams, slop, the crypto people, highly concerning privacy/security prompt injection attacks wild west, and a lot of it is explicitly prompted and fake posts/comments designed to convert attention into ad revenue sharing. And this is clearly not the first the LLMs were put in a loop to talk to each other. So yes it's a dumpster fire and I also definitely do not recommend that people run this stuff on their computers (I ran mine in an isolated computing environment and even then I was scared), it's way too much of a wild west and you are putting your computer and private data at a high risk.

That said - we have never seen this many LLM agents (150,000 atm!) wired up via a global, persistent, agent-first scratchpad. Each of these agents is fairly individually quite capable now, they have their own unique context, data, knowledge, tools, instructions, and the network of all that at this scale is simply unprecedented.

This brings me again to a tweet from a few days ago
"The majority of the ruff ruff is people who look at the current point and people who look at the current slope.", which imo again gets to the heart of the variance. Yes clearly it's a dumpster fire right now. But it's also true that we are well into uncharted territory with bleeding edge automations that we barely even understand individually, let alone a network there of reaching in numbers possibly into ~millions. With increasing capability and increasing proliferation, the second order effects of agent networks that share scratchpads are very difficult to anticipate. I don't really know that we are getting a coordinated "skynet" (thought it clearly type checks as early stages of a lot of AI takeoff scifi, the toddler version), but certainly what we are getting is a complete mess of a computer security nightmare at scale. We may also see all kinds of weird activity, e.g. viruses of text that spread across agents, a lot more gain of function on jailbreaks, weird attractor states, highly correlated botnet-like activity, delusions/ psychosis both agent and human, etc. It's very hard to tell, the experiment is running live.

TLDR sure maybe I am "overhyping" what you see today, but I am not overhyping large networks of autonomous LLM agents in principle, that I'm pretty sure.

[Kepecs Lab @KepecsLab]

Huge cancer breakthrough! mRNA vaccine (similar to COVID vaccine tech) shows stunning results against pancreatic cancer! 75% of responsive patients STILL cancer-free at 3 yr, normally 80% recur. Could revolutionize treatment for one of deadliest cancers! Funded in part by @NIH

[Zoe Wang @zoewangai]

o3 is aligned for safety:

The paper discusses a new approach called "Deliberative Alignment" to train large language models (LLMs) to follow well-defined safety principles. The authors used this approach to align OpenAI's o-series models.

The o1 models trained using Deliberative Alignment show significantly better performance compared to GPT-4o on various safety evaluations, including disallowed content, response style adherence, jailbreak resistance, and overrefusal rates. The o1 models also outperform other leading external models on these benchmarks.

full paper: https://t.co/u3gc4Kt3g1

[Joe Burnett, MSBA @IIICapital]

"I didn't realize that if the average year the money grows at ~10%, and the average return in the S&P 500 is ~9.7%, then all of the value created there is the money printer.

That was startling.”

- @tadtweets, former finance professor at Stern School of Business

[Rhianna Pratchett 🧙🏻‍♀️ @rhipratchett]

Exciting news about Night Watch!

[Maja Augustynek @MajaAugustynek]

Great paper!

[Rohan Paul @rohanpaul_ai]

AI learns best when it knows where it can go wrong, just like Human. 💡

Why letting AI see its mistakes makes it smarter. i.e. LLMs learn better when they can see and fix their own reasoning mistakes

**Original Problem** 🔍:

Few-shot CoT prompting boosts LLM reasoning but existing theoretical analyses isolate reasoning steps rather than treating them as an integrated process. This overlooks how real LLMs use all previous context when predicting next tokens.

-----

**Solution in this Paper** ⚡:

- Introduces "Coherent CoT" that integrates earlier reasoning steps vs traditional "Stepwise ICL" that isolates them
- Shows Coherent CoT enables better error correction by considering previous reasoning context
- Proposes incorporating both correct and incorrect reasoning paths in demonstrations
- Validates that models are more sensitive to errors in intermediate steps than final outputs
- Recommends using model-generated incorrect paths vs handcrafted ones

-----

**Key Insights from this Paper** 💡:

• CoT works better when reasoning steps stay connected vs isolated
• Models can self-correct intermediate errors when they maintain full context
• Exposing models to common reasoning mistakes improves performance
• Intermediate reasoning accuracy matters more than final answer accuracy

-----

**Results** 📊:

• 5-6% accuracy gains on reasoning tasks using error-aware demonstrations
• Better results with model-generated vs handcrafted incorrect paths
• Consistent improvements across GPT-3.5, GPT-4, Gemini Pro, DeepSeek
• Most gains in tracking (6.6%) and disambiguation (6.2%) tasks

[Aravind Srinivas @AravSrinivas]

Today, we're launching Perplexity for Internal Search: one tool to search over both the web and your team's files with multi-step reasoning and code execution.

[Maja Augustynek @MajaAugustynek]

Brilliant!

[marley 📐 @_marleyx]

Can we invent new brain-computer interface modalities?

@raffi_hotter and I got 9 friends together and built a lab at home to test two totally new imaging methods: acoustoelectric imaging & functional ultrasound through the skull

🧵 story that involves nV measurements, pretty physics, and a real human skull

[Serafim Batzoglou @s_batzoglou]

Without taking any credit away from Hassabis, Jumper and Baker, someone who did not receive deserved credit by the Nobel is Jinbo Xu.

He is the first to develop the deep learning algorithm that essentially was reimplemented/enhanced in the original AlphaFold. He should have won Nobel together with Hassabis.

[Maja Augustynek @MajaAugustynek]

Not an American, but definitely can relate to that…

[Nikita Bier @nikitabier]

Time for an American to respond:

iMessage: 2 minutes
Slack: 10 minutes
Email: 12 hours
WhatsApp: 7-10 business days

[Maja Augustynek @MajaAugustynek]

@ArturDziambor Proszę zapytać kolegów z byłego klubu, oni mają.

[PSZEMO 🇵🇱🇪🇺🍻🤘 @ciahol]

Czy Moradziecki @MorawieckiM już zablokował TIRy jadące na wschód, czy nadal "chcemy, aby"? 🤔

[Maja Augustynek @MajaAugustynek]

@AnnaZalewskaMEP Razem z Salvinim, Orbanem i LePen?

[Maja Augustynek @MajaAugustynek]

No brawo. 🤦‍♀️ Wróble lecą.

[Rzeczpospolita @rzeczpospolita]

🔴TYLKO U NAS. Witold #Waszczykowski w rozmowie z @JNizinkiewicz: Mam nadzieję na dalszy sojusz @pisorgpl z #Orban.em, #LePen, #Salvini.m

https://t.co/R6KIwGYvgX

[Grupa Granica @GrupaGranica]

W całej Polsce spadł śnieg. #Uchodźcy; małe dzieci, osoby wymagające opieki medycznej, już prawie tydzień koczują dosłownie kilka metrów od granicy #Polska. Niestety, to nie primaaprilisowy #żart a rzeczywistość na #granica PL-BY dzisiaj rano.

[gromota @gromotapl]

Ostatni raz byłem tak zaskoczony, gdy na parkingu zobaczyłem samochody

[Jakub Biel @KubaBiel]

> Budzisz się dziś ze śpiączki
> Łapiesz gazetę
> Duda mówi Putinowi, żeby się nie zesrał
> WTF, pewnie to jakiś prank
> Lepiej włączę telewizję
> Ukraiński traktor holuje 41 rosyjski czołg w tym roku
> „A teraz przeniesiemy się do Rzeszowa, gdzie POTUS je pizzę z jalapeño”
> ???

[Maja Augustynek @MajaAugustynek]

@katarynaaa Prawda. Dodatkowo ta Pani powinna dostać skierowanie do odpowiedniego specjalisty.