adaptAI Inc.的动态

Finally some real innovation in LLMs that is not about spitting out the next token better. OpenAI o1 series models actually "think" before they respond, very much like a human's internal reasoning before answering a hard question. We already started benchmarking the o1 model today and we are excited to bring this superhuman intelligence into solving complex problems in manufacturing with Aibuild. Congratulations to OpenAI, and all pioneers who dare to build something new. https://ai-build.com/ai/ https://lnkd.in/eKT_XGV6

OpenAI Nears Release of ‘Strawberry’ Model, With Reasoning Capabilities “The experience of using OpenAI’s updated AI system will differ somewhat from what people have come to expect with ChatGPT, the company’s chatbot. Before responding to a user’s prompt, the new software will pause for a matter of seconds while, behind the scenes and invisible to the user, it considers a number of related prompts and then summarizes what appears to be the best response, the person said. This technique is sometimes referred to as “chain of thought” prompting.“

Axel Springer, owner of media giants like Business Insider and Politico, has partnered with OpenAI, the minds behind ChatGPT. This raises questions: is it a counter to xAI's rise, or just a strategic move? Get more insights from our short analysis. #OpenAI #gpt4 #xai #grok #AIandJournalism

Rumors: OpenAI working on a core action model for the next big release? Given the capabilities of their #LLM by today's standards, moving to the adiacency of task-based operational models has the potential to create a massive category killer that goes way beyond modular agents. Maybe the path to #AGI is shorter than we think? (paywalled article) #genai #LAM #innovation

OpenAI's acquisition of Rockset is an exciting partnership and forms the basis for OpenAI’s becoming an answer engine for the enterprise. The possibility of combining real-time enterprise content indexing with high quality reasoning (e.g. ChatGPT) is the holy grail for enterprise nowadays. I see stiff competition arising for anyone who is building LLMs to answer business-specific questions. ?? How do you think enterprises will decide what AI platform to go with??

Great interview here of OpenAI CTO Mira Murati by Joanna Stern of The Wall Street Journal, focusing on the status of OpenAI text-to-video generator Sora. Hits on big topics of interest to me, including how much processing power it takes to generate a video, the sources of data used in the Sora model and the anticipated costs to use Sora for generating video. Fuzzy answers to several questions indicate that we're still in early stages of Sora's development. In general, Murati is promising of broader access to tool "later this year," but it's clear that Sora remains a work in progress. Even Google acknowledges that Sora is ahead of the curve, but one has to ask whether OpenAI announced these capabilities a bit too early. #IntelligentVideo #ArtificialIntelligence #Sora

Thanks, James, for this enlightening update. Indeed, events happening at OpenAI are exciting and underline certain aggressive dynamics of the AI sector. Still, while Apple's exit is symbolic, the fact is that OpenAI could raise $6.5 billion with strong validation from investors despite the challenges. The forecasted loss of $5 billion for 2024 underlines the enormous consumption needed to develop leading-edge AI, mainly by talent acquisition and computing power. Of course, this development raises more important questions about the sustainability of current AI business models and what it says about the future of the industry. The interesting thing, though, will be how other tech giants will respond to this and eventually shape the future of AI innovation and funding strategies. #AIIndustry #TechInvestment #OpenAI #FutureTech

LET'S "PUT ALL TRAFFIC LIGHTS ON GREEN." The example mr. #Harari (big fan btw) addresses, did occur. Under guidance of OpenAI's Red Team, an agent GPT, or #LAM (Large Action Model, which sounds better than 'agent', right?) as Jesse Lyu for rabbit inc. coined. Which will be useful ordering a pizza too, or -think about that for a moment- an ambulance, might you or a loved one need one. OpenAI's Red Team, pushes to test to see what could go wrong with OpenAI's tech. To learn what safety measures need to be built in. In this specific example (as far as I know), not just ChatGPT-4 was deployed. It was an experimental #agentGPT version built on model 4, that was allowed to -under supervision- develop its own tactics and store the results, to learn from those. In this test case, the model managed to go where you and I can go online, solving a CAPTCHA. That said; #OpenAI's latest product - #01 - is a similarly working construction... An 'agent' or a 'LAM' or an 'agent swarm'. You could try doing 'bad' things with it. As you could with just anything. But why should you? Since the first 'agents' were issued, a lot of people built similar constructions with open source LLMs. Anything can be built by anyone opting for open source solutions. As Rutger Bregman concluded: Most people just want to do the good/ right/ decent thing. And frustrated teenagers in desperate need to make a statement, in the US can still easily get their hands on semi-automatic rifles and try to settle a new record school shooting instead of deploying AI-agents. They'll score more press attention that way too...

OpenAI just skyrocketed its value to $80 billion. Thanks to a strategic play with Thrive Capital, OpenAI's crew is cashing in, marking a jump from last year's $29 billion. This isn't just growth; it's a seismic shift. Remember the buzz around ChatGPT in November 2022? That was OpenAI grabbing the spotlight. With Sam Altman steering, they've banked over $1.6 billion in 2023. Backing this titan is Microsoft, with a $10 billion vote of confidence, fueling OpenAI to push the envelope further. And with Sora that is now turning text into video, it feels like OpenAI isn't stopping at innovation, they're leading it and redefining the AI landscape always with an eye on what's next. PS. If you're as obsessed with AI and business as I am, let's connect! Drop me a message and let's geek out together.

Can OpenAI charge enough to cover its costs? And does generative AI keep its buzz when there's a real price tag attached? "Roughly 10 million ChatGPT users pay the company a $20 monthly fee, according to the documents. OpenAI expects to raise that price by $2 by the end of the year, and will aggressively raise it to $44 over the next five years, the documents said." https://lnkd.in/g_ZiEtbe