Nvidia reports that according to their data Nvidia H200s are profitable for AI companies. $1 in H200 cost can generate $7 in revenue over 4 years serving Meta Llama 3. This means if an AI company buys a $40k H200, the AI company can make $280k in AI over 4 years. The Nvidia H200 has …
OpenAI GPT 4O is free to use (limited queries) but has 5 times the usage limits with paid access. It is faster and better than the competing AI models. It clearly shows that the foundational models will be free. Fast and free was discussed by Sam Altman as a goal in his All-in- podcast interview. …
In 2022, there were projections predict that we will have exhausted the stock of low-quality language data by 2030 to 2050, high-quality language data before 2026, and vision data by 2030 to 2060. This might slow down ML progress. However, the conclusions rely on the unrealistic assumptions that current trends in ML data usage and …
Meta has put Meta AI front and center in the search bars for Facebook and Instagram that have served as key tools to find content from particular friends or creators. This is what the All-in-Podcast guys called scorched Earth for AI. This means almost no one will pay the $20 per month or other subscriptions …
Ate-a-Pi has a weird claim that Mark Zuckerberg does not believe in AI compute scaling delivering better performance. There is also a claim about energy constraints for data centers and AI chips maximizing on performance per watt. Zuckerberg says that the exponential curves are still scaling. He says no ones how long that scaling continues. …
Grok-1.5V is competitive with existing frontier multimodal models in a number of domains, ranging from multi-disciplinary reasoning to understanding documents, science diagrams, charts, screenshots, and photographs. Grok has capabilities in understanding our physical world. Grok outperforms its peers in our new RealWorldQA benchmark that measures real-world spatial understanding. For all datasets below, they evaluate Grok …
I have used the new open source Meta Llama3 AI model. It can generate images and you can edit the images by asking for changes to previously generated images. The system is faster and easier to use than many of the other image generators. It is pretty good at math and coding. It was able …
Youtuber, Internet of bugs, broke down the Devin Upwork video frame by frame. He shows what Devin was supposed to do, what it actually managed to do instead, and how bad a job of that it did. Devin was not given the right instructions by the Cognition Employee. It was supposed to write instructions for …
Llama 3 is Meta’s latest generation of models that has state-of-the art performance and efficiency for openly available LLMs. Meta AI is available online for free. The small 7B model beats Mistral 7B and Gemma 7B. The 70B beats Claude 3 Sonnet (closed source Anthropic model) and competes against Gemini Pro 1.5 (closed source model …
Reka Core is a new LLM that is competitive with OpenAI GPT4 and Anthropic Claude 3. Reka Core can identify an African Elephant from a picture by looking at its ears. It was a model created by a team of twenty people. Reka, a California-based AI startup, appears to be ready to go head-to-head with …
The various Claude 3 models are ranking very closely to the OpenAI GPT 4 AI models. Claude is able to understand images. I uploaded the picture of a SpaceX Starship refueling in space. Claude Sonnet described its understanding as follows. The image depicts two futuristic-looking spacecraft or rockets in space, seemingly in orbit around the …