OpenAI has announced the latest iteration of its popular chatbot ChatGPT called GPT-4. This update to the company's artificial intelligence language model has already announced a new string of customers as well, including Morgan Stanley's
Morgan Stanley announced Tuesday that is starting to roll out GPT-4 to aid its thousands of financial advisors as they gather insights from data to better serve clients. The bank has been testing the AI tool with 300 advisors and plans to launch the service widely in the coming months, Jeff McMillan, head of analytics, data and innovation at the firm's wealth management division, told CNBC.
McMillan said GPT-4 is like "have our chief strategy officer sitting next to you when you're on the phone with a client."
OpenAI said that its new chatbot is already powering Microsoft's
The company said in a blog post that the new model is "more creative and collaborative than ever before," and "can solver difficult problems with greater accuracy."
OpenAI said the distinction between GPT-4 and its predecessor GPT-3.5 is "subtle" in casual conversation, but it "exhibit human-level performance on various professional and academic benchmarks," evidenced by it passing a simulated bar exam with a score around the top 10% of text takers. To compare, GPT-3.5's score on the same exam with around the bottom 10%.
The company said the difference between the two models "comes out when the complexity of the tesk reaches a sufficient threshold -- GPT-4 is more reliable, creative and able to handle much more nuanced instructions than GPT-3.5."
GPT-4 can also accept both visual and written prompts --a notable improvement from GPT-3.5 --though it can only respond via text. The program also has similiar limitations as earlier GPT models, including responding with inaccurate information (Open AI says it "hallucinates" facts) and has the capacity to generate harmful outputs.
"GPT-4 generally lacks knowledge of events that have occurred after the vast majority of its data cuts off (September 2021), and does not learn from its experience," OpenAI wrote. "It can sometimes make simple reasoning errors which do not seem to comport with competence across so many domains, or be overly gullible in accepting obvious false statements from a user. And sometimes it can fail at hard problems the same way humans do, such as introducing security vulnerabilities into code it produces."
OpenAI does note that GPT-4 has made improvements in these particular areas compared to earlier models, with GPT-4 being 82% less likely overall to respond to requests for "disallowed" content compared to GPT-3.5 and responds to sensitive requests, like medical advice, in-line with OpenAI's policies 29% more often.
Beyond integrations with software providers, the new model is available to the general public through ChatGPT Plus, OpenAI's chatbot subscription, for $20 per month.