CJ Fairfield’s Post

Anthropic recently made several announcements, including an upgraded Claude 3.5 Sonnet and a new model called Claude 3.5 Haiku. One announcement that seems to be generating the most buzz is a new “computer use” capability within Claude 3.5 Sonnet, which one expert says outpaces similar capabilities in OpenAI’s ChatGPT. Check out this article from MES Computing for five things to know about Anthropic Claude’s new computer use feature:

ChatGPT 4.0 is the latest version of OpenAI's language model, released on May 21, 2024. It has improved accuracy, can process text, voice, and images, and offers customization options for more personalized responses. These improvements make conversations with ChatGPT 4.0 more reliable, interactive, and tailored to individual needs.   What additional benefits can it bring to procurement?  

"Building a large language model requires analyzing patterns across a huge trove of human-written text. All of that computing takes a lot of electricity and generates a lot of heat. To keep it cool on hot days, data centers need to pump in water — often to a cooling tower outside its warehouse-sized buildings. In a paper due to be published later this year, Ren’s team estimates ChatGPT gulps up 500 milliliters of water (close to what’s in a 16-ounce water bottle) every time you ask it a series of between 5 to 50 prompts or questions. The range varies depending on where its servers are located and the season. The estimate includes indirect water usage that the companies don’t measure — such as to cool power plants that supply the data centers with electricity." https://lnkd.in/gcdTYG8U

I asked ChatGPT to summarize the article about Joseph Weizenbaum - the computer scientist who created the first chat bot, ELIZA, in 1966. Here is the summary: https://lnkd.in/gNPaYVAC and here is the full article: https://lnkd.in/grbkDdBy

Question for ML/AI experts: I'm stuck after doing a lot of research on this topic. Can anyone please help? T5 (enc-dec) vs Llama (dec-only) architecture. I see FlanT5 models mentioned a lot in AI/ML research papers for specific niche tasks. But why aren't these encoder-decoder architectures talked about much in the open-source community? It looks like the focus, activity and support is mostly on decoder-only models. For example, there's not much support (or ecosystem) for fine-tuning (or inference) of T5 models, which are encoder-decoder. Is this because of ChatGPT's popularity? Are people mainly interested in chat-like AI now? The more I read, the more confused I get. The papers I read last week used T5 as basis. But there is silence on T5 from OSS community. Why isn't there much interest or support for T5? As opposed to the other. Am I missing something? Context : We need to choose a base model for a big training project (expensive) for a specific use case. And we are stuck b/w choosing the llama/mistral models or the flan-t5 models. The papers we read all have flan t5 as basis. And we are worried if we will go the "wrong" way by choosing flan t5 now. Can someone help? Would appreciate it very much!

📣 Announcing AIDE has become the first human-level AI agent for data science! AIDE outperforms half of human data scientists on a wide range of Kaggle competitions, surpassing conventional AutoML, LangChain agents, and ChatGPT with human assistance. Weco AI provides a cloud-hosted version of AIDE. They also understand that expert users may want to run it locally to leverage their own computing resources, ensure better data privacy, and achieve more customization. Therefore, AIDE will be open-sourced in the coming weeks at here. This repository currently showcases AIDE's solutions to over 60 Kaggle competitions. 👉 https://lnkd.in/d68ZNvwT #machinelearning

Anthropic just dropped a bomb! 💣 The AI race is getting more exciting by the day. With the latest update, Claude just released 'Computer Use' which can now use a computer and perform tasks on websites with just a few prompts. The demo video shows it being used for coding, but the possibilities are endless. Claude, ChatGPT, and V0 have been competing closely in AI development race, but after using all three, I’d say V0 is leading the pack (probably because they're leveraging Vercel apps in their training data). But, this new feature from Claude takes things to a whole new level. What platforms have you been using? Could this be the first glimpse of AGI?

Exciting Feature Updates Announced at OpenAI DevDay 1️⃣ Realtime API – Enable the creation of advanced speech voice agents, similar to ChatGPT’s enhanced voice capabilities. 2️⃣ Prompt Caching – Automatically cache common input prompt prefixes to reduce costs and improve latency. 3️⃣ Model Distillation – Fine-tune smaller models using outputs generated by larger models for efficient performance. 4️⃣ Vision Fine-tuning – Fine-tune vision models with image and text pairs to tailor them for specific use cases. 5️⃣ New Playground Feature – Leverage AI in the playground to automatically generate system prompts and output schemas! #OpenAIDevDay #AIUpdates #RealtimeAPI #SpeechAgents #PromptCaching #ModelDistillation #VisionFinetuning #AIPlayground #MachineLearning #AIInnovation #GenerativeAI #VoiceAI #AIModelTuning #AIDevelopment #AIEngineering #TechInnovation Umar Iftikhar

🛠️ Did you know you can build a #ChatGPT #DataPipeline with RisingWave Stream Processor and Astra Vector Search? 🌟 Enter the exciting brave new world of #GenAI with Mary Grygleski and Karin Wolok ✨. Catch their #OSACon 2023 session to get started: https://lnkd.in/dqkMPJXj #opensource #AI

The series of yarns on 'ChatGPT Doctrine' offers insights and guidelines for this revolutionary technology https://ed.gr/eg8ka #AIInsights #ChatGPTSeries