(Besides the fact that you don’t need technical skills to do it)

One week ago, on Mar 14, OpenAI released GPT-4. I’ll admit to being one of the users who rushed to pay my $20/month to play with it in ChatGPT Plus and I’m impressed. Let me take this opportunity to congratulate OpenAI on a tremendous achievement in democratizing AI. Bravo!

Today, Google released Bard, giving the public access to its alternative to GPT-4, LaMDA, for free via a waitlist. Another bravo!

What I’m far less impressed with is some of the online chatter on the subject. Too many hot takes seem to be coming from people who’ve never used it themselves and have only heard about GPT-4 from their cousin’s daughter’s yoga teacher’s prize cabbage.

Prize cabbage, as seen in Madrid. No idea what I’m looking at, but I love it. Photo by the author.

I could join the fray and write all about how GPT-4 and LaMDA aren’t perfect and they still makes plenty of factual errors (though they’re less gaudy than the mistakes we’ve been poking fun at for the last few months) but here’s what I’ll say instead:

Firstly, never trust anything you haven’t tested thoroughly. Secondly…

Stop reading about GPT-4 and LaMDA and go spend that time getting your hands dirty with them.

I feel personally compelled to point you in the direction of self-improvement: right now, GPT-4 and LaMDA are the most cutting edge AI productivity tools released to the general public, so please invest some time in becoming a more AI-informed citizen — stop reading about them and go spend the time getting your hands dirty with them. (Feel free to abandon this article too, I’ll survive.)

That’s how you’ll know what they are and whether they’re useful to you. I promise you that you don’t need any special training to interact with them, all you do is you type your text in a text box, just like using a text messaging app:

• To try GPT-4, you just sign up for ChatGPT Plus and off you go. (Costs $20/month.)
• To try LaMDA, you just sign up for Bard and off you go. (Free, but there might a waitlist.)

Dive right in! You can’t do any harm with these things if you keep your (un-fact-checked) output to yourself. Avoid putting confidential/sensitive info in there, though.

The fact that there’s two of them means my advice doubles: try them both! Copy-paste from one to the other. Feel free to drop the same text into both and see what comes out.

Instead of reading what people have to say about them, try them both out and see which one suits your needs better.

When you’re done, scroll down for the second half of my article.

This is the first output from running the same query I used for my viral ChatGPT post in December 2022. As a writer, I’d use zero of this output verbatim. It’s certainly not in my style. (If I’m ever that chipper and so vomitously overdescriptive, please put me on an urgent caffeine detox.) But the output is compelling enough for me to have a productive argument with it, which fuels better writing. Feel free to compare the GPT-4 output above with its predecessor’s here.

Back from trying it out? Good. Now we’re ready for an intelligent discussion! Hopefully you’re simultaneously impressed (what a useful tool) and unimpressed (it’s certainly not everything those social media posts claim it is).

Large Language Models (LLMs)

LaMDA and GPT-4 are to large language models (LLMs) what the latest laptops are to computers. As we saw with the jump in usefulness from ChatGPT to today’s ChatGPT Plus, the tech is evolving quickly. Honestly, it’s my hope that the first half of this article ages as badly as a poem praising the first Macbook. Innovation never sleeps, so you can expect that the future will bring us far better LLMs than GPT-4 and LaMDA.

LLM technology has enormous potential and, no matter who you are, it would be unwise to ignore it.

The important thing to know is that LLM technology has enormous potential and, no matter who you are, it would be a dodgy career move to ignore it, which is all the more reason for you to have a hands-on sense of what it’s about.

Even CEOs will be using it

How often does a technology come along where the C-suite not only needs to be championing its adoption (by other people) but should personally start using it immediately?

One of the most audacious things about the LLM revolution is that almost no one’s work is spared a potential productivity boost. Yes, this technology has its place on the CEO’s own laptop, right next to email and Solitaire.

Does Freecell count as Solitaire? Source: Wikipedia

I’m suddenly reminded of a factoid I heard about how secretaries in the 1970s feared they’d be replaced by robots… but were instead replaced by their own bosses. Previously accustomed to dictating all their correspondence, managers now felt that clacking away on a computer was, unlike a typewriter, no longer beneath them and absorbed many of their secretaries’ duties.

Image by the author. AI-generated, naturally.

That’s the kind of revolution we’re about to see, folks. Not job replacement by machines but rather an unprecedented spike in individual productivity, which raises opportunities and problems for society as a whole. I’ve written more about that here.

2022: A productivity revolution

The year that changed the way we work

kozyrkov.medium.com

That said, insisting on org-wide adoption without investing a few minutes to try it yourself seems especially obnoxious in this particular case. Don’t be that CEO. Instead, fiddle with an LLMs until it disappoints you. That’s how you get a sober sense of its potential and its limitations.

You shouldn’t be sitting this one out and waiting for it to blow over. Maybe GPT-4 lasts, maybe it’s a mayfly, but LLMs in general are definitely worth trying out, even if you’re allergic to technology.

You shouldn’t be sitting this one out and waiting for it to blow over.

Regardless of the underlying LLM, the user experience (UX) is as straightforward as they come: you log in, you type in whatever text you like, and you press ENTER. Even if a better LLM comes out tomorrow, getting started as soon as possible is time well spent.

There’s a lot to say on this topic, so I’ll continue in a short series. In my next article (subscribe to stay tuned), I’ll explain a hidden gotcha that’ll prevent many companies from adopting LLMs despite all the excitement.

Thanks for reading! How about a YouTube course?

If you had fun here and you’re looking for an entire applied AI course designed to be fun for beginners and experts alike, here’s the one I made for your amusement:

Enjoy the course on YouTube here.

P.S. Have you ever tried hitting the clap button here on Medium more than once to see what happens? ❤️

Looking for hands-on ML/AI tutorials?

Here are some of my favorite 10 minute walkthroughs:

• Vertex AI
• AutoML
• AI notebooks
• ML for tabular data
• Text classification
• Image classification
• Video classification

Liked the author? Connect with Cassie Kozyrkov

Let’s be friends! You can find me on Twitter, YouTube, Substack, and LinkedIn. Interested in having me speak at your event? Use this form to get in touch.