Conversations With AI: Generating Text with Artificial Intelligence and Deep Reinforcement Learning, a Conversation With Jeff Dean, Jeremy Dean, and Jeff Dean
The underlying technology at the center of chatgppt is not new. It is a version of an AI model called GPT-3 that generates text based on patterns it digested from huge quantities of text gathered from the web. That model, which is available as a commercial API for programmers, has already shown that it can answer questions and generate text very well some of the time. It took a certain prompt to get the service to respond in a certain way.
Reddy, CEO of Abacus. AI, which develops tools for coders who use artificial intelligence, was charmed by ChatGPT’s ability to answer requests for definitions of love or creative new cocktail recipes. Her company is already exploring how to use ChatGPT to help write technical documents. “We have tested it, and it works great,” she says.
OpenAI has not released full details on how it gave its text generation software a naturalistic new interface, but the company shared some information in a blog post. The model was pushed to provide better answers to the questions by using a form of reinforcement learning, which was fed as training data.
A professor at the MIT who works on Artificial Intelligence and language says the system may expand the pool of people willing to use it. He said that if the thing was presented to you in a familiar interface, it would cause you to apply a mental model that is used to interacting with humans.
Late last year, I attended an event hosted by Google to celebrate its AI advances. The company’s domain in New York’s Chelsea neighborhood now extends literally onto the Hudson River, and about a hundred of us gathered in a pierside exhibition space to watch scripted presentations from executives and demos of the latest advances. Speaking remotely from the West Coast, the company’s high priest of computation, Jeff Dean, promised “a hopeful vision for the future.”
There’s something strange happening in the world of artificial intelligence. In the early part of this century, the field burst out of its inertia by the innovation of deep learning. Our applications were made more useful by the approach to artificial intelligence, and many were powered by it. We’ve spent a dozen years in this AI springtime. But in the past year or so there has been a dramatic aftershock to that earthquake as a sudden profusion of mind-bending generative models have appeared.
Answers to those questions aren’t clear right now. One thing is. Even as current giants are laying off employees, the granting of open access to these models has begun to affect the tech sector in a positive way. The next big paradigm isn’t the metaverse, it’s the new wave of artificial intelligence content engines. The golden age of moving tasks from paper to PC application was in the 1980s. In the 1990s, you could make a quick fortune by shifting those desktop products to online. The movement was mobile a decade later. In the next decade, there is a shift to building with generative artificial intelligence. Thousands of startup will come up with their plans this year due to the access to the data in those systems. The cost of churning out generic copy will go to zero. By the end of the decade, video-generation systems are expected to be the dominant force in TikTok. They may not be anywhere as good as the innovative creations of talented human beings, but the robots will quantitatively dominate.
Openai has created the new enterprise tier called Foundry, which allows companies to run the latest GPT-3.50 model with dedicated compute designed for large workloads. Over time, Spiegel says, it will use the data gathered from the chatbot to inform its broader artificial intelligence efforts. While My AI is basic to start, it’s the beginning of what Spiegel sees as a major investment area for Snap and, more importantly, a future in which we’re all talking to AI like it’s a person.
“The big idea is that in addition to talking to our friends and family every day, we’re going to talk to AI every day,” he says. “And this is something we’re well positioned to do as a messaging service.”
That distinction may mean that it isn’t as bad for snap. As Bing’s implementation of OpenAI’s tech has shown, the large language models (LLMs) underpinning these chatbots can confidently give wrong answers, or hallucinations, that are problematic in the context of search. They can be emotionally cruel if they are too toyed with. It has kept larger players like Google and Meta from releasing competing products to the public.
Snap is in a different place. It has a deceivingly large and young user base, but its business is struggling. My Artificial Intelligence is expected to boost the company’s subscriber numbers in the short term, but eventually, it could open up new ways for the company to make money, though Spiegel is cagey about his plans.