Not a vacuum… At its first in-person event on Monday, OpenAI presented GPT-4 Turbo, an update to its popular large language model. The company said Turbo will have lots of bells and whistles, including a half dozen preset voices and the option for customers to create customized ChatGPT-powered bots. Picture: task-specific bots that can be listed for sale on a soon-to-launch GPT Store (like the App Store, but for AI). OpenAI said ChatGPT has 100M weekly users, but many use it for free.
How OpenAI makes $$: In addition to charging $20/month for its ChatGPT Plus subscription, OpenAI charges developers for access to its model.
Turbo opportunity: AI servers are incredibly pricey to run, and the GPT store could create an opp to boost revenue.
Due for an overhaul: Researchers said that Turbo’s predecessor, ChatGPT-4, had become less capable (aka: dumber) over time.
Turbochargin’ competition… OpenAI said that 92% of Fortune 500 companies use its platform, but it’s far from the only game in AI-ville. Elon Musk recently said that his xAI would release its own ChatGPT rival. Musk said the bot, called “Grok,” would eventually be available to paid X subscribers, and that it’d answer “spicy” questions. Also in the mix (* takes deep breath *): Google’s Bard, Anthropic’s Claude, Meta’s Meta AI, and scores of up-and-comers like Aleph Alpha, Inflection AI, and Cohere.
A crowded track can lead to crashes… As companies rush to get ahead in the AI race, collisions may be inevitable. Researchers said chatbots regularly make factual errors (when tested, Bard “hallucinated” 27% of the time). Experts say potential AI harm includes people being falsely ID’d as criminal suspects and AI-powered autonomous cars not accurately spotting pedestrians. Officials are paying attention. Last week, President Biden signed an exec order on AI safety.