Jing Guo

Some belated news... Excited to share that I was accepted into the @Cohere_Labs Catalyst Grant Program last month, and receive the Tinker Research Grant from @thinkymachines this month! Huge thanks to both organizations! Looking forward to using these grants to solve problems and advance research in AI + math!

@itsreallyvivek Thank you! DMed.

@guojing0 sure, just hit me up w your questions in my dms

grateful to share that i’ve been accepted into the MATS program in the empirical ai track, and will be joining in Berkeley this September. coming right after the Anthropic AI Safety Fellowship in London, this feels like a quiet, surreal back-to-back moment. trying to hold it with gratitude more than noise. there’s a lot to learn, a lot to prove, and a lot of work ahead

@yacineMTB I see, thank you for the clarifications. I forgot if it’s you or someone else, also mentioned in the past that EPFL does really good ML research (without relying on too much compute). Out of curiosity, which GPU(s) would you recommend, how about 5060 Ti 16 GB?

@guojing0 i'm talking about their research. it's often really great results with a single gpu

if you're doing AI research at all; I recommend doing the "ETH zurich" route Train models that use a single GPU. Make sure that it takes less than a minute to train models. Pufferlib is a great example. The more models you train the more you learn

@itsreallyvivek Thank you! Much appreciated

Let's not forget the real person who inspired the recent AI proofs, Paul Erdős. I was lucky to see him lecture while a grad student. In 2013, we celebrated his centenary. blog.computationalcomplexity.org/2013/03/a-cent… And I highly recommend Paul Hoffman's biography "The Man Who Loved Only Numbers".

Major difference in my mind: - an engineer, given a problem, invents and tries multiple solutions and stops when the solution is good enough. The goal is product innovation and shipping. - a scientist asks new questions, proposes various new solutions, compares them (sometimes with old ones), and writes about it. The methodology must be sound or else peers will sneer. The goal is scientific breakthroughs and technological progress. Both can be called "researchers". Many people can do both: these are activities, not identities. Importantly, most product innovations are built on scientific breakthroughs and technological innovations that happened 2, 5, 10, or 20 years earlier.

@apartovi For students who already graduated, the Neo Residency is the only track?

It means the world to me when I hear something like this from the students we've mentored. Thank you Jason!!! 🥹❤️

@Yuchenj_UW The title has a long history, but here’s how we decided to bring it to OpenAI:

@SuJinyan6 Hope you enjoyed your time at Meta! Back to PhDing?

I finished my internship at Meta today, and it feels so hard to say goodbye. Over the past 6 months, the world has changed tremendously, it feels harder and harder to find places and people that are willing to devote their time and energy just to see the growth of juniors. The job market has also become less friendly to new grads. I tried to blind myself with model safety and alignment (the easier part). When I stop looking away, it keeps flooding back to me: Isn't the biggest safety problem people's job security and dignity? And shouldn't we align humans first before aligning the model? When jobs become simply the way to survive, and when humans are priced, compared and evaluated simply by their "value for money" relative to AI, I can't image a future like this.

I was reminded recently of an article by David Brooks in The Atlantic, titled "You Might Be a Late Bloomer." Two paragraphs that have stuck with me: "We have a notion that the happiest people are those who have aimed their life toward some goal and then attained it, like winning a championship trophy or achieving renown. But the best moments of life can be found within the lifelong learning or quest itself. It's doing something so fulfilling that the work is its own reward. 'Effort is the one thing that gives meaning to life,' the Stanford psychologist Carol Dweck once wrote. 'Effort means you care about something.' 'The secret of life is to have a task, something you devote your entire life to, something you bring everything to, every minute of the day for the rest of your life,' the sculptor Henry Moore once told the poet Donald Hall. 'And the most important thing is—it must be something you cannot possibly do.'"

@Right_Inverse True.

@guojing0 maybe, but I don't think that can be relied upon to continue on the scale of, say, decades

it is difficult to not be disheartened by the recent progress of AI in math. I am starting a 4-year postdoc in the fall. it is not difficult to imagine that I won't be able to recognize the profession of "mathematics" by the end of it.

When Nature reached out to use the graph I created (using GPT) to illustrate the new (dis)proof of the unit-distance problem, I reached out to Will Sawin to see if he had other suggestions. So here is a slight modification that bounds the complex norm of the points.

@emilyzsh Could you please take a look at DM, thank you!

ps we are hiring :)

some personal news! it is rare to find a company that is genuinely novel and singular -- couldn't be more excited to be building the legal experience of the future with @ryanjdaniels @jsarihan & the incredible Crosby team

@dlouapre It’s Will Sawin, so…

In only 1 day, a human mathematician improved the bound on the Erdős conjecture counter-example. GPT proof implies δ at least 6.10^{-38} Will Sawin improved it to 0.014 I guess it means humans are still 10^35 times better than AI, right?

@lintangsutawika Good for you! Helped with CMU PhD admission? :)

Working with EleutherAI has been such a transformational experience for me; it has opened doors that would otherwise remain shut for a person such as myself. Consider applying if you are interested in AI research but come from a rather unusual background!

@haoyue_bai @UWMadisonCS @WIDiscovery @rdnowak Congrats, way to go!!!

Celebrating a very special milestone in my life: my PhD hooding and graduation ceremonies from @UWMadisonCS @WIDiscovery🎓 I am deeply grateful to Professor Robert Nowak (@rdnowak) for being an extraordinary mentor, role model, and educator. I have always been inspired by his breadth of knowledge, charisma, and vision. More importantly, his unwavering support and individualized mentorship have shaped not only my research journey, but also the kind of researcher and person I aspire to become. I would also like to sincerely thank @OpenAI for its generous research funding support throughout my PhD studies. This support enabled me to pursue ambitious research ideas and grow tremendously as a researcher. A heartfelt thank you as well to all my excellent collaborators, labmates, internship mentors, friends, and family members for your years of support, encouragement, companionship, and belief in me. This journey would not have been possible without you. Looking forward to the next chapter ahead! ✨

Following up on the suggestion from Will Sawin, here is an illustration of the new configurations that disprove Erdos' unit distance conjecture (made with the help of ChatGPT 5.5 Thinking).