Intercom on Product: How ChatGPT Changed Everything

[embedded content]In a recent episode, our Director of Machine Learning, Fergal Reid, shed light on the latest breakthroughs in neural network technology. We discussed DALL-E, GPT-3, and if the hype around AI is just that or if there was something to it. He told us things were starting to pick up steam. And just like that, here we are again.

ChatGPT, OpenAI's prototype artificial intelligence chatbot, launched last week and it's been making the rounds of the halls of the internet, inspiring rave reactions from diehard techno-positivists to tech-skeptics perpetual. The bot is powered by GPT-3.5, a text-generating AI, and according to OpenAI, it can generate text in a dialog format, which "allows answering follow-up questions, admitting mistakes, challenging incorrect premises and reject inappropriate requests."

Although it is still early to see it applied to real uses, it is undoubtedly very promising. In fact, for Fergal Reid, the shift in capacity we've seen over the past year suggests this space could be "as big as the internet." And that's why we decided to bring you a special episode on these latest developments in the world of AI, what they mean and if it's time to apply it in real scenarios such as customer support .

Here are some of our favorite takeaways from the conversation:

By pushing the scale and training these models with more and more data, these robots began to exhibit qualitative changes like learning abstract concepts without supervised learning. Currently ChatGPT works best on problems where it is easy to validate the answer or creative contexts where there is no correct answer. Although we see significantly better reasoning skills from these models, they still have problems with hallucinations - if they don't know something, they make it up. If you prompt these models with the "let's think step by step" prompt, the accuracy rates increase and you get better inputs than just having it instantly give the answer. Our technology interfaces are gradually becoming more conversational, and we're only just beginning to see the quality of natural language understanding getting good enough to unlock them. There are many exciting applications of this technology in support, such as agent augmentation, but more work needs to be done before it can be deployed.

If you like our discussion, check out other episodes of our podcast. You can follow on iTunes, Spotify, YouTube or grab the RSS feed in the reader of your choice. The following is a slightly edited transcript of the episode.

The great beginnings of ChatGPT

Des Traynor: Hi, Fergal.

Fergal Reid: Hey guys. How is it going? Thank you for seeing me again.

Des Traynor: Good. It's good to see you again. We had you just five weeks ago on the podcast to talk about what's going on with AI. And you're back because other things happened.

Fergal Reid: It's been a busy five weeks.

Des Traynor: It's been a busy five weeks and seven days. Seven days ago, it was Wednesday, November 30, and I received an email with an open beta invite for something called ChatGPT. What happened?

"It went viral, it went crazy, and everyone was really excited"

Fergal Reid: What happened? So that's an interesting question. OpenAI released their most recent machine learning system, the AI ​​system, and they released it very publicly, and that was ChatGPT. And it's quite similar to their current offering, GPT-3, GPT-3.5, but it was packaged differently, you didn't need to put a credit card in it, and I think everyone just see that "Wow, there's been a huge capacity change here recently. And it went viral, it went crazy, and everybody got really excited. And around the same time, they got released their most recent GPT-3.5 model, like davinci-003, which does a lot of the same things, and it's maybe a little less good for saying, "Hey, I'm a big language... .