Twitter Beğeni Hilesi

I am attending AAAI 2024 & co-organising a workshop on fact verification. This is an extended shared task from one of the papers aclanthology.org/2023.acl-long.… aaai.org/aaai-conferenc… If it interests you, you can attend on 26 Feb, Room no. 202 Amit Sheth Prof. (Dr.) Amitava Das AAAI

Congratulations Dr. Bihorac!

This week, we are holding the 2nd Summar AI Camp for High School Students (it is a revision of aiisc.ai/safari/). Here is part of the team present at the end of today (my PhD student group is majority women): photos.app.goo.gl/eVVq7G72TkJtLh…

One clear, sad conclusion that we have to draw from our AIW study (arxiv.org/abs/2406.02061) is that all current SOTA open-weight LLMs that claim strong performance (eg Command-R+, Mistral, Dbrx-Instruct, Llama 3, Qwen, etc) are in fact seriously flawed in simple basic reasoning.

🤔 Unfortunately, there isnt yet a robust definition of what constitutes simple reasoning. Reasoning by approximation is precisely defined and whether or not that translates to simple/non-simple reasoning seems entirely subjective (determined on a paper-to-paper basis).

It is a highly coveted and challenging skill. Today, I heard someone call the successful training attempt of a large language model wizardry. I'd find it hard not to be smug if I were not a muggle in a world of muggles -- You could write a seven-book series about such smugness.

Mouth watering Gentleman's singles final on Sunday 😋

⚡️ Excited to share that I am starting an AI+Education company called Eureka Labs. The announcement: --- We are Eureka Labs and we are building a new kind of school that is AI native. How can we approach an ideal experience for learning something new? For example, in the case

I cannot wait

Based. Im trying to learn to use this word -- and, I think it fits in this context.

Yummy 😋

This is in relation to investors not getting their "get rich quickly" wishes. The bubble around exploring the technology's utility, is just starting and now the science can actually build on preliminary successes and progress. Good riddance to the buzz-filled smoke and mirrors.

The number of times a day I end up on this search page is dumbfounding.

This -- about training language models from scratch is seriously misunderstood, even after all the resources available out there that make the difficulties/trade-offs plentifully clear.

Space in my opinion is the most natural confluence of multiple scientific pursuits. Makes me happy to see the 🇮🇳 prime minister endorsing space research and progress 🙂

This is peak Djokovic today. Astonishing tennis! 🎾

It seems simple -- The insinuation is that these actions are deliberately (maliciously/without just cause) geared towards "blacklisting" X as a bad place to advertise, and hence is construed as "less than legal"

Roger Pensorse is 93 today, and claims that this number is not a big deal 🙂 Its good to get reminded of the foolishness of blowing things out of proportion sometimes, and instead stay steady and focused despite difficulties. Happy birthday good Sir.

Isnt it better to simply say statistical compression machines have these faults? To invent new things you need things akin to proof systems, that derive new statements from given statements, and compression machines have a role to play within such systems - e.g., neurosymbolic AI

Since I've been into tennis (2006-7 onwards), this is the first year that one of fed, nadal, or nole haven't won a grandslam (the official start of this stat is 2003). Absolutely mind boggling longevity.