Sabitlenmiş Tweet
Aban Mandal
109 posts
Aban Mandal
@aban_mandal
Exited founder. Building in AI x Material Science. PhD researcher @ UW
Seattle Katılım Haziran 2025
447 Takip Edilen35 Takipçiler
@dr_alphalyrae Was at a housewarming party in SF, realized it was more of a networking event than a fun party! Is it common in SF/Bay Area for parties to be mostly networking events for founders and VC’s
English
This was the highlight of my SF trip over the weekend.
As aptly put in the billboard, AI becomes useful when it fits into how work actually gets done rather than generation of content.
@sam_ridd_ho thanks for being a wonderful SF host
English
@VideoTutor_io @ericzhu Are you sure teachers haven’t evolved? It’s bold of you to say goodbye to teachers.
English
VideoTutor is out. Goodbye Teachers.
We developed an AI that monitors, teaches, and builds a unique learning experience for every student.
For years technology has evolved. Teachers have not. It’s time for a change.
English
I think the tension is less “Use of AI in research” and more about fit. AI can clearly help with parts of research, but ultra AI pro folks are pushing it beyond its current limits and treating fluent output as understanding. Research still needs deep reading, judgment, and accountability.
English
I think the concern should go beyond a simple “mistake.” The bigger issue is that AI-written drafts can turn into technically dense word soup that looks plausible at a glance. The focus should be on preventing those cases, which is still downstream of actually reading the paper carefully.
English
My issue with "you just need to read the paper" crowd: if I upload a paper with "TODO: Tom, please edit" in a caption nothing happens; so it feels weird to asymmetrically punish a mistake because AI was involved (and yes, mistakes can happen).
Mathieu@miniapeur
Although I generally feel something should be done about arXiv, I can understand the backlash given the severity of the proposed punishment. Could we at least institute a points-based system, like for a driving licence? A direct one-year ban, plus losing the ability to share unpublished work on arXiv, seems too harsh. That said, failing to notice that a reference is incorrect is different from submitting a paper that contains an AI prompt, which could literally be detected by reading the said paper.
English
@Quasilocal Seriously, I think the AI bro’s are going too far with this. If AI is the only entity that can read a paper, research is doomed already.
English
Hot take: If you don't even want to read the paper once, maybe don't put your name on it.
Michael 英泉 Eisen@mbeisen
The funniest thing about #arxivgate is that it’s going to drive everyone to use AIs to ensure their papers are in compliance because AIs are literally the only entity with the patience to read an entire paper.
English
@ziv_ravid IMO it’s the first author/ first co author’s and the corresponding author’s responsibility.
English
A few more thoughts: I agree with the concern that Arxiv will become full of millions of AI-slop papers written by agents. **But** this is not the answer. Not only because banning an author for life is not fair (it might that you write only some sections and ofcourse that you can't check all the references one by one), but also because this is not the real problem. AI is writing our papers at this point, and the only question is at what level. If I ask the model to rephrase a sentence and it keeps, " Sure, I will do it for you", it doesn't mean anything about the quality of the paper. The same as we didn't punish papers with grammar errors in the past. Moreover, I can create entire fake papers that will pass this bar from one prompt, and soon, we will have systems that will catch these mistakes automatically. At the end, we need to come up with a semi-automatic system that will check for correctness without the question of whether it was generated by AI (the question of whether we want humans to be responsible for each paper is a different story)
Ravid Shwartz Ziv@ziv_ravid
We must think about how to handle the fact that LLMs can generate papers without any human intervention, but this is not the way. Also, I want to see the arXiv ban senior PIs who upload 40 papers a year.
English
Kind funny how established researchers just to further their AI tools would go on to diminish the sanctity of research. How can you not know what’s in the paper that you are publishing.
Chirag Gupta@seekergupta
@tdietterich @arxiv Did you also ban articles that used internet for research in 2005?
English
@tunguz How would you read the paper then, if you don’t have access to it?
English
Here is another big issue with "checking all the references by all authors on the paper": most of the times many of the references are behind paywall that many collaborators don't have an easy access to. This is especially the case for small independent researchers with no institutional backing.
English
@octonion Absolutely, I wanna meet these people and ask why they don’t want rigorous standards for research.
English
The backlash against arXiv is a bit odd. All they're asking is that you read your papers before submitting them.
English
@lpachter Isn’t it the responsibility of the corresponding author as well to ensure they know what they are pushing out. IMO the onus is more on the professor.
English
Upon submission of a preprint, arXiv should conduct three hour oral exams for each author. Families should be invited so they can witness humiliation of the authors should the opportunity arise. Professors should be exempt from the exams (but invited for champagne afterwards).
English
@tdietterich @arxiv Crazy how people are trashing on @arxiv for implementing checks and balances on AI use, make me wonder how much research they have actually themselves done.
English
Attention @arxiv authors: Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents were generated. 1/
English
Finally some one said it! SF tech twitter would put out things like “xyz is dead…. abc just released….that would revolutionize the sector..” often being ignorant that these tools persisted for reasons that automation alone cannot solve.
Jaya Gupta@JayaGup10
Spent a few days in New York catching up with folks: 1.) Few have heard of Mythos 2.) While people have heard of Claude Code, few have tried Cowork or knows what a skill is 3.) Many F500 non tech companies here aren’t getting any value out of AI because of politics 4.) Too many companies have too much AI slop esp in product roles, removing AI now 5.) Lot of “Gemini” shops since it’s bundled 6.) The same people firing people because of AI are also the same people that still want their decks sent to them as a PDF / printed on desk This technology will take a longer to diffuse than SF tech twitter thinks
English
@JayaGup10 Finally some one said it! SF tech twitter would put out things like “xyz is dead…. abc just released….that would revolutionize the sector..” often being ignorant that these tools persisted for reasons that automation alone cannot solve.
English
Spent a few days in New York catching up with folks:
1.) Few have heard of Mythos
2.) While people have heard of Claude Code, few have tried Cowork or knows what a skill is
3.) Many F500 non tech companies here aren’t getting any value out of AI because of politics
4.) Too many companies have too much AI slop esp in product roles, removing AI now
5.) Lot of “Gemini” shops since it’s bundled
6.) The same people firing people because of AI are also the same people that still want their decks sent to them as a PDF / printed on desk
This technology will take a longer to diffuse than SF tech twitter thinks
English
Getting your AI to write an entire Systematic review for your in 1 day is the first sign that one is a 0/10 researcher.
People need to stop solve the field of research like they did for domains like coding.
Faheem Ullah@Faheem_uh
PhD Students – How to write a systematic literature review in 1 day? A systematic literature review takes 4-6 months. You can reduce this time. 🎯 Here is how you can write it in 1 hour. 1️⃣ Go to gatsbi.com 2️⃣ Select Gatsbi reviewer from the drop-down menu 3️⃣ Enter the topic of your literature review 4️⃣ Gatsbi will generate an outline for review 5️⃣ If you are OK with it, click on write manuscript. 6️⃣ Gatsbi will write the literature review for you. 👉 The literature review contains the following parts ✓ Title ✓ Abstract ✓ Introduction ✓ Methodology ✓ Results ✓ Discussion ✓ Conclusion ✓ References 👉 This polished paper also contains ➝ Diagrams ➝ Tables ➝ Equations ➝ Graphs Once the draft is ready, you can humanize the text. Once humanized, you can download it in the following formats. ↳ MS Word ↳ Latex ↳ Markdown After downloading, you can make any changes you want. In addition to Gatsbi Reviewer, you can also use: → Gatsbi Innovator: Generate ideas before writing → Gatsbi Writer: Write research papers 🎗️Try Gatsbi today for free: gatsbi.com
English
That's exactly the inflection point we're approaching.
Like LLMs Biology models will capture structure humans missed.
But here's where it gets wild: with LLMs we understood the training data. With biology the model might surface a mechanism that's real and actionable before any human can explain why it works.
Recursion is already finding targets in pathways nobody was looking at. That's not optimization. That's discovery in a black box.
English
@aban_mandal The jump from token prediction to codon prediction is wild when you put it that way. Curious how long before the models start surfacing biology we haven't even thought to look for.
English
They trained GPT on text from the internet. Now they training biology models on data from our DNA.
ARK's multiomics flywheel thesis: more data → better AI → cheaper drugs + actual cures. The first CRISPR cure is already priced at $2.2M — and payers are covering it.
The next decade in biotech will make the last decade in software look slow.
ARK Invest@ARKInvest
The convergence of AI and biology is not a future possibility. It's underway. Which companies do we see building the future of healthcare? Read @Shea_ARK's The Multiomics–AI Flywheel: Why ARK Believes AI And Biology Are Converging To Reshape Healthcare arkinvestmentmanagement.substack.com/p/the-multiomi…
English