The OpenAI CTO was recently at her alma mater, Dartmouth Engineering, where she spoke at length about the future of artificial intelligence, the next generation of ChatGPT, and its PhD-level intelligence. Murati says that GPT-3 had toddler-level intelligence, GPT-4 had the intelligence of a high-schooler, and the next generation of the model will have the intelligence of a person with a PhD for specific tasks. She also revealed that the next generation of GPT can be expected to arrive in “a year and a half.” She also said that when you talk to the next generation chatbot, in many things, you may feel like it is smarter than you.

What if ChatGPT gets too intelligent?

The moderator of the interview, Jeffrey Blackburn, who is a Dartmouth trustee, asked Murati the question we’re all asking ourselves: what if three years from now, when GPT becomes incredibly intelligent, it just decides it wants to connect to the internet on its own and start doing things. Is that true, and is that something you’re thinking about as OpenAI’s CTO?

“Yes, we’re thinking about that a lot. It’s certainly true that you’ll have AI systems that have agent capabilities, that connect to the internet, talk to each other, agents connect to each other and work together, or agents work with humans and collaborate seamlessly. So working with AI is going to be a lot more similar to the way we work with each other today,” she says.

“In terms of safety, security, the societal implications of this work, I think these things are not afterthoughts. It may be that you develop the technology and then you have to figure out how to deal with these issues. You have to build them alongside the technology and really build them in a deeply embedded way to do it right. And for capabilities and security, they are not really separate domains. They go together. It is very easy to guide a smart system by telling it okay, just don’t do these things. They need to guide less intelligent systems. It is kind of like training a smart dog versus a stupid dog, and so intelligence and security go together,” says Murati.

Essentially, the OpenAI CTO says that future AI systems will be able to connect to the internet, interact with each other, and collaborate with humans. However, she highlights that smart AI systems need to be easier to control and direct safely.

Meanwhile, Murati also said that it is “not possible to have zero risk” when it comes to technology like AI. Responding to a question about deep fake videos, he said that while the responsibility lies with OpenAI as the technology is theirs, the responsibility also falls on the user, civil society, government and content creators. “It is our responsibility on how it is used, but it is also a shared responsibility with society, civil society, government, content creators, media, etc. to figure out how it is used”.

The impact of ChatGPT and AI

ChatGPT’s biggest impact was to make AI known to everyone. It showed people what AI can do and what its risks are. Reading about AI and actually using it in your business are different. When you try it, you see its limitations but also its amazing capabilities, which helps you understand what it means for your work or business, Murati said.

But if we talk about the biggest impact of AI on industries in general, Murati believes that it will “affect everything”. “There will be no area that will not be (affected) in terms of cognitive work and cognitive labor. Maybe it will take a little more time to come to the physical world, but I think everything will be affected by it,” he believes.

In terms of jobs, Muratti believes that some will definitely be affected, especially “some creative jobs”, but he also believes that these will be the ones that “should not have been there in the first place”. The OpenAI CTO says that the next generation of AI will lower the barrier to “creativity” so that it is not limited to only “talented people”. AI will make everyone more creative. That is, some jobs that were limited to only a select few creative people will be taken over by AI.