Chat-GPT had 1 million users by the fifth day after its release. OpenAI, the creators of Chat-GPT, put the technology on the internet and did not expect this speed of growth. Here's what happened next ... and how far OpenAI CEO Sam Altman expects the technology will go.
“We woke up one morning, and like any other product, we put it out there and we’re like, we hope people like it,” said Altman during a recent interview at Elevate. “People did like it. And by two o’clock that afternoon, it was clear something big was about to happen. It wasn’t clear how big.”
Altman said it is very tempting to “anthropomorphize” this technology, but reminds people this is a tool. Not a “creature.”
Altman also pioneered ventures like the early social media app Loopt and led the start-up incubator Y Combinator. His lead at OpenAI, initially a non-profit created with Elon Musk and others, has positioned him as a key influencer in AI development.
Humanity should be as proud of this as the moon landing, he said. After so much scientific and technological progress, people have figured out a computer algorithm with no tricks and no sleight of hand that can actually learn, despite debate over what that really means. And it gets predictively better with skill.
“Those two things taken together is a profound thing for what will happen in the coming decade for humanity,” he said.
With GPT-4, people are building entire companies around it. Advertising agencies are using the technology to create and build campaigns.
Reprise developed products that use the GPT framework to make content such as ad-copy variations and meta data. Cross media used AI as assistant to internal teams and to ideate and copy variations and in the creation of new keywords and copy.
Ciceron is still beta testing, but at the MediaPost Search & Performance Insider Summit said Chat GPT 4 does an excellent job at unstructured data out of CRM to learn how to prompt the AI and deliver insights that normalize language. Ciceron uses it for new persona and journey profiles.
Early on, Altman stated in a blog post that it is very possible to predict very early on in the process of training the model how smart it will be. In other words, predicting the intelligence of the AI before it’s trained, and that it follows close to a natural law based on computer processing power and the size of the neural network and training data to determine the capabilities of the model.
“But can we predict the qualitative new things, the new capabilities that didn’t exist in GPT-4 that do exist in future versions like GPT-5,” he said. “I think we will look back at this like we were all living through one of the most important periods of human discovery.”
Altman also spoke about “AI workers” that join and exist in society along with human workers. They will become a “fully remote virtual employee,” team members. Another way to think about it, he said, is when the AI system can semi-autonomously or helping humans, discover new science. He said, if the rate of scientific progress humans makes increases by a factor of 10, 100 or 1,000, that feels different than a bunch of remote employees.
“The fact that AI can do creativity, and the fact it can use code to verify things actually gives me hope that we might have an AI that can do science,” he said.