Cellgeist

Cell (English: the fundamental unit of life) + Geist (German: mind, spirit).

[Cellgeist @cellgeist]

@noahatwhere 3. but something real please. Not your afternoon AI slop.
9. some of my best projects were side projects at the beginning.
11. sleep& exercise is the most underrated advice

[Cellgeist @cellgeist]

@BarteldsMartijn Thanks!

[Cellgeist @cellgeist]

@ATinyGreenCell I’m curios. what do you need benchling for these days ?

[Brian Armstrong @brian_armstrong]

Publicly funded science should work more like open source software.

Today a result is a PDF. The data often isn't shared, the code isn't public, and few studies ever get replicated.

Imagine every experiment as a repo. Data, code, and protocol, all versioned. Anyone could fork it, rerun it, and send a fix. Funders could reward replications as much as new papers.

AI makes this urgent. Agents are now making more discoveries than ever. The bottleneck becomes replication/validation. Then commercialization, which still requires moving atoms in the real world and meetings with regulators.

[Cellgeist @cellgeist]

@CNGI_news Just a Hedgefond with a small research & teaching arm

[Cellgeist @cellgeist]

@LayerbaseDB @iuditg Doing what exactly? And how many agents in parallel?

[Cellgeist @cellgeist]

@nsaphra But let’s talk about the message not the messenger.
Do you agree with their points or not?

[Cellgeist @cellgeist]

@natalienkhalil It would probably become more reproducible. But the biomedical research reports from OpenAI and Anthropic have been pretty disappointing so far (very preliminary stuff, not solving any important problems, etc.), as there is a human component that we cannot yet replicate in an LLM

[Cellgeist @cellgeist]

@iuditg That’s wild. What the hell are you doing?

[Cellgeist @cellgeist]

@araseb_ 💯 Absolutely! 🚀 Authenticity is KEY. 🔑

In a world full of AI, being HUMAN is your superpower. 🤖✨

[Cellgeist @cellgeist]

WHY are people scared that AI solves all existing scientific problems (math, physics, biomedicine, etc.)?

We can always create new problems. Throughout history, people have loved solving completely irrelevant ones (chess, video games, crossword puzzles, etc.).

Those are designed problems with artificial constraints, and humans adapt very well to them.

We derive tremendous satisfaction from solving them even without external benefit.

I once thought Google and later ChatGPT would kill trivia games and pub quizzes, but they had basically no effect on their popularity.

Or e.g. Chess is more popular than ever, even though machines are now 10x better than humans.

[Cellgeist @cellgeist]

It’s crazy that basically no one knows guys like @signulll or @tszzl and they could completely fake everything (achievements, stories etc.), but their tweets are so insightful that they require the intellect of someone extremely competent.

[Cellgeist @cellgeist]

I've been using Meta AI and ChatGPT on a long flight with only messaging apps available.
In theory, that's a great hack to access web and AI models via WhatsApp, but in practice, those models seem pretty bad (lots of hallucinations, struggling with web access, etc.).

Why are the WhatsApp versions lagging so far behind?

[Cellgeist @cellgeist]

@ymatias @Nature @BIDMChealth @harvardmed Respect to you guys for actually working on this for years and then publishing this in a peer-reviewed journal.
(Compared to other AI companies that announce preliminary & unvalidated 4 weeks of research as major AI breakthroughs)

[Cellgeist @cellgeist]

To those complaining: Yes, Opus is still better than Astra. But in reality this doesn’t make a difference in my opinion.

[Cellgeist @cellgeist]

Codex is by far the best bang for your buck for coding & bioinformatics projects

Based on the current @ArtificialAnlys benchmarks:

Simple tasks
- Terra: 48 score vs. Haiku: 37 score.
- But OpenAI 3.9× cheaper.

Medium tasks
- Sol: 54 score vs. Sonnet: 59 score 
- But OpenAI 8.9× cheaper.

Hard tasks
- Astra: 55 vs. Opus: 62
- But OpenAI 1.6× cheaper

Scores = Coding Agent Index.
Costs = API cost/task.

[Cellgeist @cellgeist]

@justinskycak And at the end of the day completely knocked out.

[Cellgeist @cellgeist]

@GoogleResearch Respect to you guys for actually working on this for years and then publishing this in a peer-reviewed journal.
(Compared to other AI companies that announce preliminary & unvalidated 4 weeks of research as major AI breakthroughs)