OpenAI announces an API for ChatGPT and its Whisper speech-to-text tech

ChatGPT can now be officially integrated into products and services by developers. Get ready, everyone.

Third-party developers will now be able to integrate ChatGPT into their apps and services via an API, and doing so will be significantly less expensive than using OpenAI’s existing language models. The company is updating its developer terms of service and making Whisper, its AI-powered speech-to-text model, accessible through an API.

OpenAI claims that its ChatGPT API can be used for more than just creating an AI-powered chat interface. However, it also highlights several businesses that have been utilizing it for that purpose, such as Snap, whose My AI feature was announced earlier this week. The company claims that its new GPTS 3.5-turbo model family is the “best model for many non-chat use cases.”

It is important to keep in mind that the model is probably not the same one that Bing is using. Microsoft has described the model as a “new, next-generation OpenAI large language model” that is “even faster, more accurate, and more capable” than ChatGPT and GPT-3.5. However, it should come as no surprise that the company has access to technology that the average developer would not have access to given the amount of money it has invested in OpenAI. For Bing, Microsoft is also utilizing some of its technology.

OpenAI claims that the price of 1,000 tokens is “10x cheaper than our existing GPT-3.5 models” as a result of “a series of system-wide optimizations.” Although 1,000 seems like a lot, consider that sending the API a single text fragment could cost several tokens. The text blocks that the system breaks down into tokens are sentences and words that it uses to predict what text it will output next.)

The documentation for OpenAI states, “ChatGPT is great!” takes six tokens, which its API divides into the characters “Chat,” “G,” “PT,” “is,” “great,” and “!” The organization gives a device to check the number of tokens that it’ll take to decipher a line of text and says that an overall principle of thumb is that “one token for the most part relates to ~4 characters” in English.

If developers are processing an enormous amount of data through the API, the company claims that they will also be able to obtain a dedicated instance of ChatGPT. According to its post, doing so will give you more control over the model you use, the time it takes to respond to requests, and the length of conversations with the bot.

OpenAI has also announced a new API for its speech-to-text model Whisper, although ChatGPT is likely to receive the most attention. According to the company, you can use it to translate or transcribe audio for $0.006 per minute. The Whisper model is technically open source, so you don’t have to pay anything to use it on your hardware. However, OpenAI probably has access to more powerful hardware. If you need transcription done on phones or other devices with lower power, you might want to use the API.

Additionally, OpenAI is announcing some policy modifications that it claims are the result of developer feedback. A significant one is that it has stated that it will no longer use data submitted through the API to train its models unless customers explicitly consent.

At the end of the day, it’s going from a quit framework to a select-in-one. Because some businesses have completely prohibited employees from using the technology, this change may help alleviate some concerns regarding the incorporation of proprietary data into the bot. It is a bad idea to input trade secrets if it is learning from user input because there is always the possibility that it could give that data back to someone else.

Additionally, the business claims that its “engineering team’s top priority is now the stability of production use cases.”

Even though several developers have devised workarounds for including chat services in their apps, such as by utilizing OpenAI’s standard GPT API, which has been available for some time, the launch of an official ChatGPT API appears to be the moment when the floodgates will open. Even though many businesses are developing their own AI chatbot models, the majority of developers are completely unable to do so. They can now only make use of OpenAI’s technology.

Would you allow ChatGPT to manage your connected home?

Today, you can ask Alexa to turn on the lights or Siri to tell you your bedroom’s temperature, and sometimes they will. Alternately, you might hear, “You have fifteen devices named lights; Which one do you want to have control over? or “The temperature is 53 degrees right now in Kathmandu.” But what if your voice assistant could respond to vague statements like “I’ve had a rough day;” in addition to being always accurate? what’s an effective method for loosening up?” with answers that are “intelligent”? For instance, by lowering the blinds, lowering the temperature, lowering the lights, and waiting in line for some Netflix goodies?

Alex Capecelatro, the co-founder of the home automation system, asserts that this is the potential of voice assistants based on brand-new AI language models. Using OpenAI’s ChatGPT, has already begun working on a prototype integration. In this proof-of-concept video, Capecelatro asks Josh’s assistant to open the shades, turn off the music, and tell him the weather (Josh already can control three things simultaneously). He then uses voice commands for the smart home that sound more natural, like “I’m filming a video; The voice assistant responds clumsily by increasing the room’s lighting because “it’s kind of dark in here.”

Intriguingly, AI language models could be used to parse natural language to improve smart home control. Capecelatro considers it to be the future. He continues, “We’re trying to figure out how good we can get at controlling your environment in a way that feels more natural and intuitive.”

Today, voice assistants typically require precise language and frequently misinterpret simple smart home commands with information requests, resulting in irritated and sometimes ineffective responses. When Capecelatro and Quark founder Tim Gill founded in 2015, they set out to solve this issue. No matter how you phrase the request, its eponymous voice assistant aims to be excellent at controlling connected devices.

Josh can interpret when he hears “satellites” rather than “turn on the lights” using extensive knowledge graph models and take the appropriate action. Although “Open the drapes” sounds like “Get some grapes,” Josh is aware that you do not reside in a vineyard. According to Capecelatro, “we know what you mean even when you say “turn on the goddam lights” because we spend a lot of time working under the hood to fix mishearing, work with different accents, understand imperfect sentences, and the like.”

Be the first to comment

Leave a Reply

Your email address will not be published.