5 Key Insights Into AI From the Lex Fridman Podcast
5 Key Insights Into AI From the Lex Fridman Podcast
Would you disclose your phone number to sign up to talk to a semi-autonomous agent?
Six months ago, my answer was a hard No. ChatGPT was just a fad that would go away after a few weeks, I told myself.
Wrong.
Like me, you've probably been bombarded by every influencer under the sun doing this-or-that with ChatGPT. Riding the wave of a popular topic.
I ignored the hype. Until I started listening to the Lex Fridman Podcast.
Lex explores the topic of AI in-depth by interviewing key players from the industry. That includes Sam Altman, the current CEO of OpenAI, the company behind ChatGPT.
These discussions touch on both the potential life-changing nature of ChatGPT and other AIs, as well as the deadly worst case scenarios of an out of control intelligent agent.
Slowly my skepticism gave way to curiosity, and I decided to try out ChatGPT for myself. What I found made me simultaneously surprised, scared, and excited for the future.
After immersing myself in these podcasts, here's my curated list of what you need to know about the latest AI trends.
1. ChatGPT is a superpower for developers
One thing Lex Fridman keeps saying is how useful ChatGPT is for day-to-day coding.
I didn't believe it, but decided to give it a shot anyway. At the time, I needed help with AWS sign in functionality.
How to configure Cognito to verify a user with an email link using CloudFormation?
ChatGPT replied with a very helpful template and explanation. I was impressed. So over the next few days I quizzed it on other coding topics: TypeScript, Gradle, DynamoDB, and more.
My conclusion was that in many scenarios where you'd previously use Google, ChatGPT gives you the answer faster.
Google gives you hundreds of answers, but ChatGPT gives you one. And normally it's right.
Now I use ChatGPT on an almost daily basis to answer coding questions. It's incredible how much it knows.
2. AI opens new possibilities we can't yet imagine
When I realised that ChatGPT was as good as the hype, a healthy dose of fear accompanied my excitement.
It was a fear of becoming obsolete. That my coding and content writing services would no longer be required.
But listening to Sam Altman talk on the podcast snapped me out of that mindset.
"If we can move more of the world to better jobs and work to something that can be a broader concept, not something you have to do to be able to eat, but something you do as a creative expression and a way to find fulfillment and happiness..."
AI will create opportunities that we can't even imagine right now. History is full of examples like this.
Telephone calls used to be manually connected by an operator. Yes, you literally picked up the phone and told someone the number. But from the early 1900s, the whole process was automated.
Today, nobody misses these jobs. And it's not hard to imagine what jobs telephone system improvements created: customer support, sales, emergency services, and many more.
There's no reason to believe the same won't be true for AI too.
3. AI will be much more than chatbots
If you've used ChatGPT, you're familiar with carefully crafting questions and reading its slowly scrolling replies.
The website evokes the same feeling I had visiting amazon.com in the 1990s. The beginning of a new era.
But I don't think we'll be interacting with ChatGPT in this way for long.
Lex interviewed Meta CEO Mark Zuckerberg, who explains why:
"I think every creator who you interact with will ultimately want some kind of AI that can proxy them and be something that their fans can interact with or that allows them to interact with their fans."
ChatGPT is what's known as a large language model (LLM). It’s a neural network trained on huge amounts of data scraped from the internet. Much comes from sites like Wikipedia.
In the future, maybe you'll be able to train your own personal ChatGPT LLM. You could feed it articles that you've posted or conversations that you've had in the past.
Keep it private and query experiences or knowledge that you've forgotten. Your personal interactive journal.
Make it public and have another version of you in the world. Interactions that you don't have time for can be handled by You 2.0.
Hopefully one day every business will replace their annoying chatbot with intelligent AIs which can genuinely help you out.
4. Who’s training the AIs?
Despite its name, OpenAI hasn't shared much about how ChatGPT actually works.
Eliezer Yudkowsky, AI researcher and skeptic, puts it like this on the podcast.
"We still know vastly more about the architecture of human thinking than we know about what goes on inside GPT."
We do at least know that ChatGPT is trained on data from the internet. That raises the interesting question of whether we can trust what it says.
Much of what we believe about the world is more opinion than fact. Thousands or millions of others likely hold the polar opposite viewpoint to yours.
Much of that debate is documented on the internet, ready for ChatGPT to gobble up. That means that even today it can give fairly nuanced answers to hot questions.
But what about biases of OpenAI developers themselves?
Who gets to choose the data on which to train ChatGPT?
If humanity becomes reliant on a single AI, those controlling it could take advantage by manipulating us in subtle ways we don't even notice.
Having multiple independent AIs that we can customise could be a good countermeasure to this. Sam Altman also predicts that we'll figure out how to make AIs far less biased than even the most considerate humans.
5. ChatGPT stands on the shoulders of giants
Despite the amazing advances made by the team at OpenAI with the latest ChatGPT 4, ultimately this remarkable technology is made possible by thousands of years of human progress.
Even the fact that there's an easily accessible record of knowledge available is astonishing.
Anyone who's ever edited a Wikipedia article, published something online, or added to human knowledge on the internet in some way contributes to the usefulness of these AI tools.
We can all benefit from that.
Where future versions of ChatGPT take us is just another step on the ladder of human progress.
Showing an interest and giving away a small piece of personal data during the ChatGPT sign-up process is a small sacrifice to make to join the discussion on these exraordinary new technologies.
Tom Gregory
If you enjoyed reading this post, feel free to share it with friends. If you didn’t enjoy this post, you can also share it. Or not. Totally up to you. ❤️