I spent 1000s of hours on competitive programming (proof-link: codeforces.com/profile/rizar). This makes me qualified to comment on #AlphaCode by @DeepMind
The result is nice, the benchmark will be useful, some ideas are novel. But human level is still light years away.
1/n
πΊπ¦ Dzmitry Bahdanau
588 posts
Team member at @periodiclabs.
Adjunct Prof @ McGill.
Member of Mila, Quebec AI Institute.
Stream of consciousness is my own.
Joined August 2017
- Adam deserves the award, but in Singapore everyone still uses SGD
- I am excited to open-source PipelineRL - a scalable async RL implementation with in-flight weight updates. Why wait until your bored GPUs finish all sequences? Just update the weights and continue inference! Code: github.com/ServiceNow/Pipβ¦ Blog: huggingface.co/blog/ServiceNoβ¦
- I'm excited to share with you that yesterday I became a PhD! Coming next: doing some good research science at fabulous @element_ai π¨βπ¬
- RIP AI Research Long Live AI Rat Race
- The reason to build AI was to accelerate science. At Periodic, accelerating practical science is the only thing we do. No detours. We donβt work on making funny videos. Or replacing human workers. Or chatbots that order your sushi. It feels great. Check out our open positions!Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
- I'm embarassed to admit that I have just grokked how amazing Python coroutines and asyncio are. I want to rewrite every single piece of code with threads I have every written! But the learning curve is steep. This great blog opened my eyes: tenthousandmeters.com/blog/python-beβ¦
- While the whole twitter is going nuts about ChatGPT, let me just say that the HELM paper by @StanfordCRFM and @StanfordHAI is an incredible scholarship masterpiece. Make sure all your students read it and see what good research actually looks like.
- if you think AI is going to kill us all, just try building a little multi-agent system and watch just how stupid LLMs get as their message history grows...
- You shut down your nuclear plants - you have to buy Russian gas. You don't want AI for killer drones - prepare to hide from Russian ones. Being overly virtuous and progressive in 21st century is suicide. Ukraine is a sober wake-up call. AI for Western armies? Hell yes!!
- Replying to @iclr_confmany many many thanks to @kchonyc and @Yoshua_Bengio for enabling the wildest ever start of my research career 2014 was a very special time to do deep learning, a commit that changes 50 lines of code could give you a ToT award 10 years later π²
- At this point I view NeurIPS as a mix of IMO, IOI, school essay context and debate club. Just a ranking mechanism to separate allegedly smarter kids from allegedly less smart. The only problem is the smartest kids just skip the whole ordeal and join fun startups instead.
- I work at a mission-driven early-stage startup with great culture and 1000s of GPUs. Looking for someone to: - write CUDA kernels in their sleep - burst pipeline bubbles - drive MFU curves to the roofline with expert parallelism Want to extract more FLOPs from our GPUs? plz DM!
- Replying to @DBahdanauTo sum up: AlphaCode is a great contribution, and AI for coding is a very promising direction with lots of great applications ahead. But this is not AlphaGo in terms of beating humans and not AlphaFold in terms of revolutionizing an entire field of science. We've got work to do.







