With interest in generative AI skyrocketing, Open AI he would be preparing to give a very interesting turn in his strategy. As reported The InformationSam Altman’s startup soon could release an open source alternative to GPT-4in order to enter the increasingly competitive field of language models open source.
The decision would have been made by the multiplication of open source artificial intelligence-powered tools, many of which are presented as alternatives to ChatGPT. The report does not specify the alleged characteristics of the model in question, although it indicates that it’s “unlikely” that OpenAI would make it as capable as GPT-4. And it makes sense, considering that it could undermine your business strategy.
Note that OpenAI offers GPT-4 to the general public via ChatGPT Plus, in exchange for a $20 monthly subscription; while also selling access to its AI tools through their APIs. To all this, it is important to remember the brutal cost of developing ChatGPT and keeping it operational. californians closed 2022 with losses of 540 million dollarsand it is estimated that they are spending around $700,000 a day just to support the infrastructure that powers their chatbot.
So, if OpenAI’s strategy is focused on developing and commercializing proprietary solutions, why release an open source AI? because the alternatives open source They have evolved a lot in recent months, and there are more and more options for engineers. This allows them to create their own models with a high level of sophistication and at a fraction of the cost of working on structures like GPT-4.
OpenAI would return to the world of open source with an alternative to GPT-4
As we already said, it is logical to think that OpenAI will not launch an open source model as powerful as GPT-4. Nevertheless, Nor would it make sense for it to present a much less capable variant, since it would not attract developers. Even more if we consider the advances of the community open source so far in 2023.
last march FLAME leakeda language model developed by Meta, just a week after be announced. This has allowed engineers in all parts of the world to study, experiment and evolve it into new tools. All this, without the need to spend millions of dollars in hardware and without requiring huge amounts of parameters to train them.
Added to this is the recent disclosure of a document signed by a Google engineer, warning that both Mountain View and OpenAI could lose the race to dominate AI against open source projects. Especially since both companies were getting distracted by the dispute between PalM 2 and Bard against GPT-4 and ChatGPT.
“While our models still have a slight advantage in terms of quality, the gap is closing astonishingly quickly. The open source ones are faster, more customizable, more private and, pound for pound, more capable.. They’re doing things with $100 and 13 billion parameters that we have a hard time doing with $10 million and 540 billion parameters. And they are doing it in weeks, not months,” said the expert.
For now, it is unknown when OpenAI would launch its new open source language model. Sam Altman’s guys last released their AI technologies in 2019, so it’ll be interesting to see how they return to that path. Once your AI open source is available, we can be sure that comparisons with GPT-4 will not be long in coming.
