Philippos Papathanos

@OdedRechavi Sounds interesting- will give it a go. What was your source of the thousands of proposals?

Important announcement!!!🫵💥💫 Would you have a tooth pulled if it helped your chances to get an important grant funded? Absurd question (obviously), but the situation right now is so bad funding-wise, that I bet some of you actually considered it for a second… Well, don’t get desperate - we created a new tool that might help! (keep your teeth!) I’m excited to announce that as of today we are officially releasing “QED for Grants” for everyone. What started off as an extension of our existing paper review platform, grew in the last few months to an entirely new design. We’ve been working like crazy on this, and although we have more things we want to add in the (very near) future, we decided to release our AI for grants NOW, earlier than planned. It’s not perfect, no AI is, but for the first time, when I run my own grants through @qedscience, I feel it gets the research, finds real problems, and gives me very useful feedback that I can implement before submission. It’s like sending it to 20 scientists from my domain, knowing they’ll agree to dedicate their entire week to carefully read and comment on every line. It’s very important to write your own grants yourself, it makes you think hard and you learn a lot from doing it, and q.e.d’s system is designed to preserve these positive aspects and augment them - you get feedback on your own writing, we don’t write for you!! But at the same time, a typical PI spends many months every year writing proposals and sadly only a tiny fraction gets funded, even if the ideas are good. When you are forced to submit an unreasonable amount of grants the quality of the writing drops, and rejection rates increase. Not because the essence is bad. It’s simply too competitive right now (the cuts made it so much worse) and if your proposal is not super clear and tight, and if it’s not a perfect fit for the grant you’re submitting, you’re doomed. Our grant solution is not an authoring, text-generating tool. It gives you constructive feedback on your writing (it comments on the deep things, not grammar and typos). It’s meant to help you with the questions that torment you late at night (“is this a good fit?”, “Is this novel enough?”, “Did I miss something?”). Tens of thousands of you already use q.e.d to improve your manuscripts and critically read papers, we built the grant tool by the same principles (you’ll identify many of the features that you told us you like). We’ve processed thousands of proposals, learned where things fail, where reviewers get stuck, why good ideas come out weak. We interviewed hundreds of scientists, and also experts who work in funding agencies and university research authorities, and implemented their feedback (we’re constantly looking for more feedback). Our AI is always happy to give you constructive (and polite!) critique, and it will go through your grant line-by-line, forcing you to improve clarity, flag weak points, and push the whole thing to a higher standard. We study, in scale, what gets funded and what doesn’t, and what is the perfect fit for each type of grant. So please, use it, pressure-test it, tell us where it fails, and together we’ll improve it every day to put you in the best position for actually testing your ideas in the real world. As always with q.e.d, the system is completely secured and private, and we are NOT training on your data (see the FAQ on our website). Please like, retweet, and share with your favorite colleagues! (link to the platform below in the thread👇)

@DieSquirrelDie @timothysolomon What i did and totally worked.

@timothysolomon Way too long and convoluted. Better: Every time your kid says “6,7” you enthusiastically join their chant. Double down when they’re with their friends. Manufacture situations where you *start* the “6,7” chant. MAKE. IT. CRINGEY.

If you kids 4 to 14 who keeps saying "6,7" this one is for you.

@AvivTamar1 @OdedRechavi Fool it in what way? fool it to hide the fact it was written with AI, I agree. But if you mean fool experts by polishing bad ideas into a wow proposal, we are not there yet, at least in the life sciences. If AI can help come up with a great ideas to include, I see no prob there.

@phiC32 @OdedRechavi Yes you can easily iterate to fool any (current) AI system. Integrity also hinges on accountability, but there's no accountability for unpublished grant proposals.

Announced by the ERC today (really bad news for science!!! It starts in Europe but will affect everybody) Because the REVIEWING LOAD is so bad, if your application fails you may need to wait mulitple years (>3?!?) until you can submit again! The ERC is the lifeline of the leading labs in Europe, so many research programs will be terminated (just no other sources to compensate). The letter from the ERC says “the only way is to reduce applications” (highlighted in the picture below) No!!! It's the wrong direction. Use AI to assist review. It’s not perfect (yet), but it allows more proposals to be evaluated, and it could support, not replace, human reviewers.

@KarinaYaniv @OdedRechavi which software? All the ones I've tested totally failed with tests I've made. Id like to test. Frankly, surprised by your confidence in you or software able to reliably know use of AI. Again, we need to distinguish between AI-junk and its use as super-thesaurus or grammarly.

@phiC32 @OdedRechavi After years of reviewing papers and grants you can definitely tell… But there are also softwares that can check that quite accurately

@KarinaYaniv @OdedRechavi I've been reviewing papers and grants for years and I definitely know i can't tell. You need to distinguish between text written exclusively by AI and full written text polished by AI.

@AvivTamar1 @OdedRechavi No! Integrity doesn't hinge on accountability. Accountability and transparency help to weed out those without integrity over time. But assuming higher dishonesty in grants due to lack of transparency seems unlikely to me. Regardless, is AI fueling a dishonesty tsunami? Doubt it

@AvivTamar1 @OdedRechavi The system depends scientific integrity, obviously. Are you saying that AI makes it so easy to fabricate data that scientific integrity will collapse?

@phiC32 @OdedRechavi Maybe, but no one can check the validity of the results and the grant submissions are not public as far as I'm aware.

@KarinaYaniv @OdedRechavi Plenty of proposals are fully AI-written...!? How can you know a proposal is fully AI-written? Just asking

@OdedRechavi The huge increase in application numbers in fact comes from the use of AI. Plenty of grant proposals these days are fully ChatGPT-written and are mostly bad. Yet, they need to be reviewed... The new system will at least make people think twice before they submit something crappy.

@KarinaYaniv @OdedRechavi Oh in fact - really? Based on? There already penalties for failed proposals not ranking highly. So, no the system isn't new, just harsher. Proposals in most institutions are screened and often reviewed prior to submission. No way what you describe is real.

@OdedRechavi They already have a cap of discussed proposals (44) per panel that are selected on the basis of Part 1 of proposal. So they are saying they cant even handle that?

@AvivTamar1 @OdedRechavi Grants are very much based on real results. It is often argued that the amount of these preliminary results expected now by panels and reviewers is unrealistic - to the point grants are being submitted whose planned experiments are mostly already done.

@OdedRechavi AI reviews for grants is a bad idea. You can program a loop that iterates on writing the grant until the AI likes it. Different from papers, grants are not grounded in real results, and anything goes.

A story that started before the lab moved to @EntomologyHUJI. Great to finally have this out. Challenging Y chromosome genomics and genetic engineering led by @roccodamato83, Elad and Alessia.

New preprint! 🦟🧬 We describe draupnir, the only Y-linked gene transcribed during meiosis in Anopheles gambiae and test whether its regulatory sequences can drive expression from the Y chromosome to power a sex-ratio distorter. tinyurl.com/4shvt7tj

Excellent work over many years by Lee Lamdan and many other lab members.

New preprint! 🦟🧬The key to engineering daughter-killing in mosquitoes isn't what X-linked gene you target with CRISPR. It's when during sperm development you cut it. X-poisoning in Anopheles at last. shorturl.at/T31TG

מכון וולקני אחראי על מחקר יישומי לטובת חקלאות ישראל, מסחור והפצה של תוצרי המחקר הוא יעד חשוב ומרכזי בליבת הפעילות. אני רואה בתפקיד, ובהצטרפות לשרות הציבורי, שליחות וציונות. זהו תפקיד המאפשר תרומה, הלכה למעשה לסקטור החקלאי הישראלי ולהצלחתו דרך מסחור הידע שקיים ומפותח במכון וולקני.

@Fredros_Inc Why does this have to develop into a conflict? The PIs raise the funds for student income from external grants and funders, and this depends and requires a record of published research. Papers are important for the development of all authors, and for science in general.

Over time, I’ve come to realize there’s a clear conflict of interest: by requiring postgraduate students, who are often based at external research institutions rather than the registering university, to publish so many papers in international peer-reviewed journals, the university lecturers (who typically become co-authors or even senior authors on those papers) get an easy route to boosting their own publication records. This, in turn, allows them to secure promotions and career advancement on the strength of their students’ work. I suspect that’s one of the main reasons these seemingly unreasonable publication quotas are so fiercely defended. In many places, even Master’s students are now expected to publish at least one paper before graduating. It’s exploitative, and it’s no wonder the system persists.

@Fredros_Inc That sucks. 3 papers is a big ask when calculating costs, especially if its mandated that pub is oa without funding support. Would posting on an archive and not pay for open access be work-around?

Local university in East Africa costs just under $2000/year for a post graduate program in tuition costs. But the university also requires that all PhD students must publish at least 3 papers in international peer reviewed open access journals . Such journals cost approximately $1800, meaning the candidate have to send nearly as much money (sometimes much more) to the publishers abroad as they pay to their local universities for tuition.

Our latest work with @DipteraAI out now on @biorxiv. We built an open, cheap and accessible mosquito membrane feeder 🦟⚙️. 3D-Feedies have been a game changer in our insectaries. Hope you'll give it a try: biorxiv.org/content/10.110…

The OpIE2-DsRed marker disrupts blood-feeding in female #Anopheles gambiae #mosquitoes by bending the proboscis. A tweaked OpIE2-b promoter fixes it. Check it out: biorxiv.org/content/10.110…