Over the past few months there has been so much hype about ChatGPT or what people refer to as generative AI. In fact, ChatGPT reached 1 million users in 5 days and 100 million daily active users in just two months beating popular messaging applications such as WhatsApp and Twitter.
So as part of that we thought that it would make sense to interview again Adam Cheyer, the co-founder of Siri (sold to Apple), and Viv Labs (sold to Samsung), and known as a world’s pioneer in the world of AI. You can also listen to the full podcast interview by clicking on the button at the beginning of this article.
?Show Notes: Throughout our conversation, we touched on what ChatGPT is, and why there is so much hype about it these days. Then we talked about what is unique about ChatGPT, and some of the most exciting uses cases for Chat GPT that he has come across. We also touched on who the winners (e.g. Microsoft, Google..) might be among the big tech companies. Then we discussed whether or not ChatGPT or similar generate AIs will eliminate jobs.
?Best Quotes: Here’s some of the key discussion points and best quotes from our conversation with Adam:
- On the progress made by ChatGPT:
- “I think that this last version called ChatGPT 3.5, surprised people, and even experts in this field. I was shocked. (…) In 2017, I was working with a team on that, and so I watched it, I saw the uses of it in 2018, 2019. And in 2020, ChatGPT 3.0 came out. I’m like, oh, okay. That’s getting a little bit better than I expected. And as I said last year at the end of 2022, they came out with a new version and it was surprising and shocking to me”.
- On Microsoft investing $10B into OpenAI the creator of ChatGPT:
- Microsoft didn’t just invest in OpenAI, they invested $10B into a startup (OpenAI). So it’s not just a minor investment. And now everyone, Facebook, Google are trying to capture mind share”.
- On the new paradigm shift happening with ChatGPT:
- “Now, the reason is, I say that every 10 plus years, the way we interact with computers changes. Back in 2007/2008 the smartphone came out. And again, there were a major winner and a minor winner namely, Apple and Google/Android. And the companies really became successful for that decade”.
- “And I believe, and I’ve been predicting that around 2020 to 2022, that a new interface paradigm would emerge. And that’s why I started Viv Labs, the kind of work in that space, et cetera”.
- “But I said back then that a conversational assistant would come out, and it’s not going to replace a graphical interface. It won’t replace mobile or the web, but I think it’s a major new paradigm. And whereas we fell a little bit short with Viv Labs. At Viv Labs we created a great technology, but it didn’t make the wide stream adoption. I think chatGPT broke through and will create this whole new era around conversational interface”.
- On what is unique about ChatGPT compared to previous digital assistants:
- “With ChatGPT, you can now ask an assistant pretty much anything and get an extremely reasonable and reasoned answer (…) And the second thing is that you can ask complex tasks”.
- On how ChatGPT can answer complex tasks:
- “So for example, I’m planning my 25th wedding anniversary, and I want to go for two weeks down the west coast. So I said plan and itinerary down the west coast of California for two weeks. Recommend cities where I should stop, and places to stay and things to see and do. And it’s for my 25th wedding anniversary”.
- “And that’s something that no Siri, no search engines could ever attempt to answer. And yet ChatGPT can say, here’s an itinerary. But first it said, congratulations on your 25th wedding anniversary. Here’s an itinerary that will be romantic and have the best aspects of California. And it went and it said day one and two, go here (..) And so on, and the coolest thing was that it was not trained specifically for that task. So you could ask a any complex questions and you’ll get a surprisingly interesting answer for pretty much anything”.
- On how ChatGPT can save your preferences:
- “Part of your question can be your preferences. So I could have said, I’d like to plan a trip for my 25th wedding anniversary. And then, I’m a vegetarian. I like basketball and sports. And I can enter in that information as part of my query or as context to the query”.
- “But you have to tell him, you have to tell ChatGPT these are my preferences. It’s not like he knows already your preferences, but it’s not too hard to write a program that saves your preferences. And for anything you said before, that you typed before, it can add at the front of that query anything you’ve saved as your preferences”.
- On ChatGPT being used into avatars and robots already:
- “There are already apps and Siri shortcuts and Alexa skills that just take the words, send it to ChatGPT, speak it back. I’ve also seen versions where they’ve hooked up an avatar, meaning a human-like Face, and I’ve even seen a physical robot that you can walk up to that has a face and some expressions using ChatGPT”.
- On which big tech companies will likely be the winners in the ChatGPT and generative AI race:
- “I wouldn’t put it past Microsoft. And which companies have a lot of our text and, and text interfaces? Microsoft and Google are two. Microsoft has Outlook with all my calendar and email. Google has Gmail (…) Microsoft has PowerPoint. There are a lot of text in that. Google also has Google, Google Slides and Google Docs, et cetera. So these technologies and large language models are primarily about text input, text and maybe graphics and commands output”.
- “So there’ll be perfect things to put into your email systems, into your writing tools. So Microsoft realized that and they’re going in hard with ChatGPT. They say it will be integrated in every technical, every product they have, and they have a lot of tech-based products. So I think they have a good shot”.
- “Google has many of the same kinds of tools as Microsoft, so I think they are well positioned and Google is an expert at scale. They are experts at running machine learning algorithms for the entire internet on massive amounts of computers. And what does this technology need? It’s very computational. It requires a lot of data and a lot of processing. And who’s got the most data? Who’s got the most processing? It’s Google, in my opinion”.
- On his belief that Microsoft has an early lead but that Google should lead in the Generate AI space in the long run:
- “I think that Google is maybe a year or two behind Open AI and Microsoft. And so are Facebook, Amazon and Apple. But if you’re looking from a long run perspective, who has the real DNA and the real skills to succeed in this space? I think Google has the most experience in what’s needed”.
- “So I would say for the short term, Microsoft will be the leader. Google and Facebook and others will be like, me too, me too. We’ve got stuff. But they’re playing catch up, but they want to capture that mindshare. And then I think in the long run, like two years plus, I think that Google can take the lead again”.
- On the fact that ChatGPT is the fastest adopted technology in history with 100M daily users in 2 months:
- “When ChatGPT 3.5 came out, everyone went, whoa, it’s the fastest adopted technology in history. There’s something like 300 million people signed up in the first two months, and more than 100 million daily active users. This is crazy”.
- “So why is that? It wasn’t because OpenAI had better algorithms, and it might not even have been that they had better data, but I think it was because they hired some people who were able to scale this up efficiently very, very well”.
- “I mean engineers, low level engineers who can really squeeze out the greatest amount of compute juice out of that lemon. And it’s an Art. And I think that their models were bigger and bigger than anyone had ever tried before. And it went like this, getting better and better and then zoom”.
- On the most exciting use cases he has seen with ChatGPT:
- “Marketing for instance, is an incredible use case. If you’re trying to write marketing copy, or build a website, it can write the copy almost better than you do the website”.
- “I’ll give you an example. I was trying to hire someone and I said, I have an open role. I’m looking for A, B, C, D, and I just had some features and qualifications. I said, write a job description for with these requirements, and it’s put together a full, beautiful, fleshed out job description. Better than I could have written in a few seconds”.
- “Here’s another example. So, as you know, I speak French very well, but I learned it 30 years ago. So I said, ChatGPT, I’d like to practice my French. I said let’s talk about a topic that I don’t know very well. Let’s converse back and forth, and if I make mistakes, please correct me. And it was correcting me as I made mistakes. It was like, and it’s perfect grammar”.
- On the fact that ChatGPT can rewrite its own code:
- “So I first started typing, write me a a sub-routine that will do X (..) And I go, okay, that’s good, but I want to change it so that this happens instead, and it rewrote its own code. And It wasn’t just spitting back some preexisting thing. I can interact with it, say what I wanted to change in English, and it understood enough to make the right changes”.
- “That blew my mind and there’s some evidence that by learning on code, by training its models over text, from code repositories like GitHub, it actually learned things about causality, loops, iterations, all these concepts of logic that matter for programming”.
- On the impact of ChatGPT on jobs:
- “I think every job such as lawyers or teachers, and even sports teams will find ways to use this technology to give them an advantage. And their jobs will change, but not be eliminated”.
- On how sports teams could start using ChatGPT:
- “You don’t have to be a computer scientist or data analyst. You can start trying these technologies today and then for every aspect of your marketing, audience engagement strategy, high level strategy, low level strategy if it’s sports related”.
- “Try putting in some statistics and asking questions about it. I’m curious to type in, “On Monday we’re going to play a team that plays small ball, runs really fast and has two good three pointers. What strategy should I use? Just to see what it says. But I bet it will give you some interesting viewpoints to kind of test your ideas and, and it’s very good at brainstorming situational type things, so I would try that”.
You may also like: