SAN FRANCISCO — Businesses and nonprofit groups agree on one thing after testing some of the latest advances in artificial intelligence: It’s already changing the course of their operations.
Five organizations that were among the first to gain access to GPT-4, the latest product from San Francisco startup OpenAI, said in interviews that they were reassigning employees, reorienting internal teams and reassessing their strategies in anticipation of major technology changes. part of your job.
Their experiences support the idea that, for better or worse, AI technology may very soon radically alter the daily lives of some people.
But the organizations also said the technology required an enormous amount of work to customize it to their specific needs, with employees providing daily feedback to the software to train it on terminology and methods specific to their fields, such as education or finance. OpenAI, best known for creating the ChatGPT AI chatbot, can then integrate the data from that work into its own model to potentially improve its technology.
In effect, each of the early testers is a microcosm of what others might experience as access to GPT-4 expands.
“There is a perception in the market now that you plug into these machines and they give you all the answers,” said Jeff McMillan, director of analytics, data and innovation at Morgan Stanley’s wealth management division.
That is not true, he said. He said the bank has 300 employees who spend part of their time testing their technology using GPT-4.
“We have a team of people who literally go through every response from the day before,” he said.
For Morgan Stanley, the result has been a specialized chatbot built with GPT-4 that serves as an internal investigative tool for its staff of financial advisors. McMillan said the tool is trained on not only 60,000 research reports on parts of the global economy, but also 40,000 other internal company documents, making it an expert on any financial topic a financial adviser might. want to consult.
To be sure, the early adopters of GPT-4 are not a random sample of the economy. OpenAI, which went for-profit in 2019, handpicked the organizations over the past few weeks and months.
Critics of OpenAI and its competitors allege that the AI industry has benefited of exaggeration without skepticism During the last months. OpenAI was looking for positive examples to show when it approached Khan Academy, an educational nonprofit, six months ago, founder Sal Khan said.
“The context was: we are going to work on a next generation model; we want to be able to launch it with positive use cases,” he said.
Khan Academy is best known for its YouTube videos, but since OpenAI reached out, Khan said he’s poured resources into creating Khanmigo, a chatbot tutor who’s specially trained in established teaching concepts.
“We collectively spent about 100 hours tuning the model so that it can behave as a really good tutor,” he said.
“If you look at the cost of tutoring, this could be a very, very significant issue,” Khan added. «It’s like having an amazing grad student or tutor or professor that you can start talking to right away.»
Stripe, a technology company that makes payments software and related products for businesses, said that when it gained early access to GPT-4 in January, it took 100 employees out of their regular jobs and assigned them to an internal «hackathon» where they each person spent an average week testing ideas.
Duolingo, a language-learning app, gained access to GPT-4 in the fall, and employees said CEO Luis von Ahn was so enthusiastic about it that he called a meeting for 8 a.m. the next morning and immediately switched jobs. I work for people.
“He, after that, said, ‘Give your team a spin,’” said Edwin Bodge, product manager. «Since then, we have been working very closely with GPT-4 and with the OpenAI team.»
So far, Duolingo has added a new paid subscription tier that costs $29.99 per month or $167.88 per year, which allows access to a French or Spanish chatbot. They also added an AI bot that will explain grammar concepts to you as you go through typical Duolingo lessons.
According to Bodge, the company has created prompts of 1,000 to 2,000 words for GPT-4 that drive the bots. The company would not share directions upon request.
All of the organizations that spoke to NBC News said they were proceeding with a degree of caution, given that artificial intelligence technology is very new and the potential danger is unknown. Mike Buckley, CEO of Be My Eyes, a company that creates an app for people who are blind or low vision, said he would like to get a test version of the app with GPT-4 into more hands, «but I want to be considered and sure».
“Could we roll this out more widely to the community in six to eight weeks? It’s possible, but we’re going to go where the data and use cases take us,” he said.
The company connects people with low vision with volunteers who, in a video call, can describe their surroundings to users of the application, such as the label of a product in a grocery store, directions to an airport or the words on a greeting card. . The GPT-4 version works without a volunteer at the other end because the AI describes what it «sees» with the camera.
One of the blind spokespersons for the app used it to get directions on the London Underground system, according to a video he posted to TikTok.
«We’ve tried to break it,» Buckley said, adding that his staff conducted thousands of tests. «We’ve been blasting the technology as loudly as we can for several weeks now, and we’ve been pleasantly surprised.»
He said his company hadn’t run into any security issues with GPT-4, but it has made mistakes; for example, mixing a toaster for a slow cooker on a website.