Since days OpenAI made a silent release. A new platform for developers with which customers could use their new machine learning models. Several users shared on Twitter some screenshots with an OpenAI explanatory document. It discussed how Foundry would enable future clients to leverage its AI models for their own platforms and tailor them to their needs.
OpenAI’s leaked Foundry pricing says a lot – if you know how to read it – about GPT4, The Great Implementation, a move from Generative to Productive AI, OpenAI’s safety & growth strategies, and the future of work.
Another AI-obsessive megathread on what to expect in 2023 ???? pic.twitter.com/OUfgryMQ3z
— Nathan Labenz (@labenz) February 27, 2023
It talked about the present of OpenAI with its current model and the immediate future, which goes through its DV models. Other simplementations that OpenAI itself talks about are Curie, Babbage and Ada, the simplest. Go-go models.
Today we have only one model, but the next iteration, known as GPT-4, is not a single model, but rather a series of “custom” models, trained on specific parameters and data to suit different use cases.
The future of OpenAI. An expert named Nathan Labenz explained in a thread on Twitter the dimension of this announcement that it was much more relevant than it might seem. It talked about the present of OpenAI with its current model (ChatGPT) and the immediate future, which goes through its DV models. What is DV?
davinci. This is the code name for current iterations of GPT-3, with Davinci (DV) being the most powerful. Other simplementations that OpenAI itself talks about are Curie, Babbage and Ada, the simplest. The latter are updated to October 2019: Davinci is updated to June 2021 and is the one used in ChatGPT.
models by gogó. Today we have only one model (ChatGPT, based on GPT-3), but the next iteration, known as GPT-4, is not a single model, but a series of “custom” models, trained with certain parameters and specific data to adapt to different use cases.
AI for Business. The DV with 8,000 maximum context tokens goes for $78,000 per month, and for the version with 32,000 tokens the price is $156,000 per month. Tokens are nothing more than sequences of characters that represent a comprehensible unit of input. They are usually a word or a small group of words that can be processed by the model.
That in essence allows us to outline in a much more precise way what we want from that model at all times. These models are intended for companies that can use them internally, or market them with specific solutions developed by them.
