Startup OpenAI has launched APIs for developers and enterprises to build its ChatGPT and Whisper models into apps, websites, and their products and services.
A joint statement from the groundbreaking new firm released this week, including remarks contributed by Greg Brockman, President and Co-Founder, OpenAI, confirmed that it was to give access to its cutting-edge language and speech-to-text capabilities. The statement read: “Through a series of system-wide optimisations, we’ve achieved 90% cost reduction for ChatGPT since December; we’re now passing those savings to API users.
“Developers can now use our open-source Whisper large-v2 model in the API with much faster and cost-effective results. ChatGPT API users can expect continuous model improvements and the option to choose dedicated capacity for deeper control over the models.”
The move to open with the APIs is hugely significant and points towards OpenAI’s strategy for making its money back. According to estimates, ChatGPT has accumulated over 100 million users since its release in December. However, many have wondered where it would make a return on investment.
The offering from OpenAI means firms can develop enhancements and integrations with a choice of models and ways to receive the service. But with OpenAI projected to make back just $200 million in 2023, its under pressure to capitalise on the buzz around its technology.
The statement added: “We believe that AI can provide everyone with incredible opportunities and economic empowerment. The best way to achieve that is to allow everyone to build with it. We hope the changes we announced today will lead to numerous applications everyone can benefit from.”
Introduction of ChatGPT Plus
Propelled by unprecedented global interest and looking to recoup the investments made from Microsoft, which placed a further $10 billion stake in the technology in January, it launched a pilot subscription plan ChatGPT Plus. The new tier was launched at $20 a month, with a free tier of ChatGPT remaining.
In a blog post on ChatGPT Plus’s release, the firm stated: “We love our free users and will continue to offer free access to ChatGPT. By offering this subscription pricing, we can help support free access available to as many people as possible,” the company continued. “We plan to refine and expand this offering based on your feedback and needs.”
ChatGPT and Whisper API Pricing
The ChatGPT model family launched this week is still the 3.5 Turbo model but will be ten times cheaper at $0.002 per 1k tokens. The startup has said this current “0301” model version will be supported through June 1st.
The API for the Whisper transcription model will cost developers $0.006 per minute, enable “robust” transcription in multiple languages, and provide translations into English.
Early Global Adopters of ChatGPT and Whisper
With the API launch, OpenAI has released a list of early adopters of the ChatGPT and Whisper models with some deeply impressive stats around their implementation.
- Snapchat creators Snap, for a start this week, unveiled My AI to its 750 million monthly Snapchatters.
- Global learning platform Quizlet will be introducing Q-Chat, a fully-adaptive AI tutor intended to engage its 60 million student user base, after three years of deploying GPT-3 across several use cases.
- The shopping app Instacart will use ChatGPT with its own AI program for 75,000+ of its retail partner store locations to harness product data to answer customer recipe questions.
- Shopify’s consumer app, Shop, used by 100 million shoppers, will be deployed to assist shoppers with buying recommendations by scanning millions of products across the database.
- Language learning app Speak is the fastest-growing English app in South Korea and uses Whisper for open-ended conversational practice.
Dedicated Instances for Developers
The firm also offers ‘dedicated instances’ to those looking for control and system performance. Instead of per request, Azure-run APIs with dedicated instances mean developers pay for an allocation of compute infrastructure per term reserved for service requests. Developers can enable a dedicated instance, and OpenAI has suggested that it makes economic sense for those running up to over 450 million tokens daily.
Developers will be able to make apps execute a range of tasks, such as: composing an email or other text, assembling Python code, answering questions about various documents, create a conversation agent, give the software a natural language interface, tutor for various subjects, translate language and simulate characters such as video games.
It looks certain that OpenAI is banking on building a platform on which enterprises can build businesses. To this end, the startup has focused on feedback and offered incentives to encourage take-up. Subsequently, it has made some changes to its terms such as simplifying data usage and making it clear to users on the input and output. It will also automate its pre-launch review process.
AI in Microsoft 365 and the Teams Platform
Previously we looked at how the potential of AI models meant developers could leverage API into Microsoft Teams to create a chatbot. The same article also showed how Teams is already using AI and primed to include meeting benefits like summaries, action planning and chapter creation.
Following this, Microsoft then announced integrations for Teams Premium and added AI to meetings with features such as Intelligent Recap, Live Translation and Meeting Templates.
Microsoft has been behind AI publicly and implementing product integration for some time, from introducing the DALL-E 2 model across much of Microsoft 365 ecosystem introducing the DALL-E 2 model across much of Microsoft 365 ecosystem back in the autumn of last year to recently unveiling its search engine Bing with a ChatGPT-powered extension. The ongoing strategy has landed at a point where Microsoft seem poised to grab Google’s dominance of the search market.
Satya Nadella, Chairman and CEO Microsoft, stated: “AI will fundamentally change every software category, starting with the largest category of all – search. We’re launching Bing and Edge powered by AI copilot and chat to help people get more from search and the web.”
Bing v Bard
According to Statista, Bing currently has 8.85 percent of the global search market, while Google’s market share dropped to 84.69 percent, down from more than 90% in 2022.
The Bing release seems part of a tennis match to see who can integrate a search chatbot quicker. Google also released Bard its storytelling AI model last month. The model is based on LaMDA (Language Model for Dialogue Applications) and has been seen as mirroring Microsoft’s ChatGPT activity since December. Yet, Google has been developing AI for over two decades and has around twenty sets of tools for developers looking to use open-source code and API services.