OpenAI makes an all-purpose API for its text-based AI capabilities – TipsClear

If you never want to try OpenAI vaunted machine learning toolset, it just got a lot easier. The company has released an API that allows developers to call their AI tools “virtually any English-language function”.

Basically, if you have found a task that requires understanding words in English, OpenAI wants to help automate it. The various capabilities of the GPT-3 family of natural language comprehension models are at the disposal of developers, at least if you can get in private beta. (Request access here.)

It is a little difficult to explain what it looks like in action, because it depends on what you want to do. For example, you can count on its ability to see a lot of text at once to answer a question about an article or find a relevant section. In a short demo video, OpenAI shows how it works – asking “why is bread so forgetful” on the Wikipedia entry for bread, part of an article devoted to the texture of bread and how it is made.

It is a simple, but famous AI dungeon is more complex. AI was given a bunch of D&D sourcebooks and adventures and based on the same, a journey was improved for the player based on his input. While previously this had to be accomplished using clumsy means, essentially to run a version of the model locally, the input could now be rendered via the bus API.

Basically it is a very easy, simple way to access GPT-3’s extensive language understanding and generation capabilities. This is just text, text is out.

The company wrote in a blog post, “We hope the API will reduce the bottleneck in the production of beneficial AI-powered products, resulting in tools and services that are hard to imagine.” Surely it would be difficult to imagine the AI ​​Dungeon a year or two ago.

It is also clear that harmful or abusive use of the API will be immediately banned, which the company finds more secure than releasing the model in the wild. “Since the downstream use cases of our models are difficult to predict, it naturally feels safer to release through an API and broaden access over time, rather than releasing an open source model where Access cannot be adjusted if harmful. Application, ”the post reads.

So far OpenAI has partnered with a dozen or more companies to test the API before offering it more widely. Chatbots, teaching assistants, legal research – there is no end to the applications of this kind of thing because language is used to define and record everything we do. Search Absolutely right Where AI agents are this way they are really useful willpower, however, use a bit.

Janelle Shane of AI Weirdness had already created a bot based on dog ratings, which makes and rates dogs in perpetuity, so this idea has already been taken.

Related Posts

error: Content is protected !!