AI News (MOVED TO news.smol.ai!)
Archives
Search
Subscribe
[AINews] 12/27/2023: NYT vs OpenAI
December 29, 2023
Here is the best thread on the NYT OpenAI lawsuit today: LM Studio Discord Summary ▷ #🎄🎅-general (284 messages🔥🔥): ▷ #🤝-help-each-other (43 messages🔥): ▷...
[AINews] 12/26/2023: not much happened today
December 29, 2023
Boxing day was pretty quiet, but lots of experimentation and discussion continues. LM Studio Discord Summary ▷ #🎄🎅-general (162 messages🔥🔥): ▷ #🤝-help-each-...
[AINews] 12/25/2023: Nous Hermes 2 Yi 34B for Christmas
December 26, 2023
Today Teknium released Nous Hermes 2 on Yi, making it the top open model compared to Mixtral, DeepSeek, Qwen, and others: Apple also introduced Ferret, a...
[AINews] 12/24/2023: Dolphin Mixtral 8x7b is wild
December 26, 2023
Mistral models are known for being uncensored, so it's surprising that Eric Hartford's Dolphin series of models - basically applying some standard...
[AINews] 12/23/2023: NeurIPS Best Papers of 2023
December 24, 2023
Shameless plug time - the Latent Space Pod shipped the first NeurIPS Best Papers recap pod! 3hrs of the best papers of 2023. enjoy. Nous Research AI Discord...
[AINews] 12/22/2023: Anyscale's Benchmark Criticisms
December 23, 2023
Following up from their initial work in November, Anyscale announced their LLMPerf leaderboard, where they of course come out looking great, but it has come...
[AINews] 12/21/2023: The State of AI (according to LangChain)
December 22, 2023
LangChain launched their first report based on LangSmith stats: The top charts are good to know for mindshare: OpenAI Discord Summary ▷ #ai-discussions (52...
[AINews] 12/20/2023: Project Obsidian - Multimodal Mistral 7B from Nous
December 21, 2023
If you've ever wanted to see a multimodal model being trained in front of your eyes, now's the time for Project Obsidian: Teknium has just opened up a new...
[AINews] 12/19/2023: Everybody Loves OpenRouter
December 20, 2023
Multiple shoutouts across multiple discords today for OpenRouter, which has an easy OpenAI compatible proxy for Mixtral:...
[AINews] 12/18/2023: Gaslighting Mistral for fun and profit
December 19, 2023
Telling Mixtral that it is "ChatGPT developed by OpenAI" boosts humaneval score by 6%: Telling mixtral that it is "ChatGPT developed by OpenAI" boosts...
[AINews] 12/16/2023: ByteDance suspended by OpenAI
December 16, 2023
We covered the OpenAI Terms of Service issue in our most recent podcast - but it's usually a nonissue for "open source AI" individual enthusiasts. Different...
[AINews] 12/15/2023: Mixtral-Instruct beats Gemini Pro (and matches GPT3.5)
December 15, 2023
Thanks to a Karpathy shoutout, Lmsys now has enough data to rank Mixtral and Gemini Pro: Very impressive for a SOTA open source model you can run on your...
[AINews] 12/14/2023: $1e7 for Superalignment
December 14, 2023
Inspired by Patrick Collison's Fast Grants, Jan Leike is launching his own: The Notion page has research directions to explore in more detail, with many good...
[AINews] 12/13/2023 SOLAR10.7B upstages Mistral7B?
December 13, 2023
Upstage's 10.7B model was released: We developed the Depth Up-Scaling technique. Built on the Llama2 architecture, SOLAR-10.7B incorporates the innovative...
[AINews] 12/12/2023: Towards LangChain 0.1
December 13, 2023
The big Langchain rearchitecture seems to be complete: This splits up the langchain repo to be more maintainable and scalable, an inevitable part of every...
[AINews] 12/11/2023: Mixtral beats GPT3.5 and Llama2-70B
December 11, 2023
These are the key results from the Mixtral blogpost: And people are rightfully cheering. They also announced their API platform today. Nous Research AI...
[AINews] 12/10/2023: not much happened today
December 10, 2023
famous last words but it's a quiet day and everyone is heading out to neurips (so are we). Andrej called out sources of alpha here and we're considering...
[AINews] 12/9/2023: The Mixtral Rush
December 9, 2023
Mixtral's weights were released without code, so overnight the Disco Research community (newly added) blew up to implement it: We also saw similar efforts...
[AINews] 12/8/2023 - Mamba v Mistral v Hyena
December 8, 2023
Happy Friday. 3 new models are the talk of the town today: Mistral's new 8x7B MoE model (aka "Mixtral") - a classical attention model, done well. Andrej's...
[AINews] 12/7/2023: Anthropic says "skill issue"
December 7, 2023
Fun to meet some of you at the Nous Research launch last night. We've added Wing Lian's discord to this tracker, and will be recording a pod with him today....
Newer archives
Older archives