Summary:
OpenAI unveiled several significant upgrades and new offerings for its popular AI systems during its first ever developer conference on Monday. The announcements highlight OpenAI's rapid progress in developing more advanced AI to stay competitive against rivals like Anthropic, Google and Meta.
The company revealed the latest iteration of its natural language model, GPT-4 Turbo. This new version contains knowledge of events up until April 2023, giving it far more up-to-date information than previous GPT-4 versions which only had data up to September 2021.
GPT-4 Turbo also massively expands the amount of context the model can take in, with a capacity to process prompts equivalent to over 300 pages of text. This enables more complex, nuanced queries according to OpenAI. For developers, GPT-4 Turbo reduces costs with input tokens priced at $0.01 per 1000 tokens compared to $0.03 for regular GPT-4.
In a major move, OpenAI is opening up the ability for anyone to create customized chatbots powered by its models without needing to code. The company introduced a platform that allows users to train chatbots through simple conversations with ChatGPT. People can add a name, description, custom instructions, conversation starters, files like Code Interpreter, specific capabilities, actions, and DALL-E 3 generated profile photos.
This provides a new level of personalization and specificity for ChatGPT, which up until now only offered custom enterprise and business versions. Over two million developers building tools with OpenAI's API can also now customize ChatGPT for their own applications and products.
To distribute these personalized chatbots, OpenAI announced an "AI app store" where creators can publish their bots for public use and potentially earn money based on usage. The company said it will highlight the most practical and entertaining chatbots in categories like productivity, education and gaming.
OpenAI also combined several AI capabilities into one ChatGPT interface, including text, image generation, document uploading and summarization as well as browsing and PDF search. The integration provides a streamlined, all-in-one experience that matches competitors like Anthropic.
Other significant announcements include a new "Copyright Shield" initiative, where OpenAI will legally defend customers and cover costs if they face copyright lawsuits over its AI systems' content generation. This echoes similar protection policies introduced by Google and Microsoft.
In additional major news, OpenAI revealed a new 128k context GPT-4 Turbo model at one-third the price of previous versions. This means GPT-4 is now cheaper, faster and can process prompts as long as an entire book.
For developers, OpenAI announced new Assistants and Code Interpreter APIs, a GPT-4 Turbo Vision API, an updated Dall-E 3 API for image generation, more realistic text-to-speech, a new Whisper V3 model and reduced rate limits to lower costs. This suite of new offerings signals OpenAI's push to advance its AI capabilities and tools for wider utility and accessibility.
The upgrades come as OpenAI aims to maintain the meteoric popularity of ChatGPT, which it revealed now has 100 million weekly active users. Since launching ChatGPT's API in March, OpenAI also has over 2 million developers building tools and products, up from just 28,000 in January 2022.
Overall, OpenAI's barrage of announcements underscores its priority of rapidly evolving its AI systems to provide greater value amid rising competition. With tech rivals like Google, Microsoft and Anthropic pouring resources into generative AI, OpenAI is making strategic moves to advance its technology while intelligently monetizing its platforms.
