Published on

Copilot is So Good

Authors

I cannot stop talking about Copilot and the-likes. It feels like AI is finally reaching a point of practicality for the average person and it could change everything really really quickly.

There's this idea of a "Singularity" where AI can improve on itself without human supervision and taking off into exponential improvement. I'd like to see that happen in my lifetime.

Singularity Chart

But even if the "AGI" doesn't pan out, I'm still excited about the tools that are here today!


First, I want to talk about GitHub Copilot. It is well worth the $100 they're charging per year. I've been using it for what feels like a year now and it has been the most helpful pair programmer I've ever had.

Some of the things I've done with Copilot:

  • Creating fake data in a CSV, JSON, etc. format
  • Writing code for my side projects
  • Writing code for work
  • Learning new programming languages
  • Solving bugs
  • Writing blog posts
  • Writing documentation
  • Writing tests
  • Writing new and complex code
  • Creating bash one-liners

It has saved me many hours of work by now and I'm confident others feel the same. I barely use StackOverflow with Copilot, every library's documentation is at the tip of my tongue, and it scans your codebase to suggest code from other files. It's outstanding.

In the beginning I was quite skeptical because I had heard of the security flaws it may introduce into your codebase but I've never had any issues with it. All of the code it produces is very readable and is on-par with the code that I'd write 85% of the time.

Pixelmator ML Tools

I've also been using Pixelmator Pro for photo editing as of recent. It has a few AI tools that are really cool. The only one I've gotten a lot of use out of is the "ML Crop" tool. It's able to frame your photo quickly and accurately.

They've also got:

  • ML Denoise
  • ML Super Resolution
  • Remove Background

and more, but I haven't used them yet. I imagine tools like these will provide lots of value to end-users today and in the future. Content tools and code-hosting platforms are the first to adopt these tools but I'm sure they'll be everywhere soon. When I heard that students are using ChatGPT to write essays, I couldn't help but laugh– it's only a matter of time.

I'm an Artist Now

When Stable Diffusion came out I spent at least 48 hours trying to run it on my AMD GPU. I was able to get it working for a few weeks but for seemingly no reason it stopped. But...

While I had it running, I was able to create some remarkable images. I have seen previous generations of AI art and it was always very blurry and abstract. With Stable Diffusion, I was able to create images that looked crisp and artistic. If you're interested, here are some of my favorites: imgur album. Use them for whatever you want.

I will admit, I only keep about 10% of the pictures it generates but I can produce 1 every 20 seconds. It's bizarre to think what we're capable of today. It's also bizarre to think about how much of it will last far into the future. I'm not sure.

Once again, this tool is great for ideas and inspiration. Stable Diffusion often takes some liberties which can be quite inspiring for your own work. If you're willing to learn the software it can even be a full-fledged art tool. I think prompt engineering will become a valuable skill in the future and being able to stitch that output together with other tools. Being able to train custom models will be valuable as well.

AlphaCode's Coding Competition

If you haven't heard, AlphaCode is a new AI by DeepMind that can write code. Here's an excerpt from this blog post:

As part of DeepMind’s mission to solve intelligence, we created a system called AlphaCode that writes computer programs at a competitive level. AlphaCode achieved an estimated rank within the top 54% of participants in programming competitions by solving new problems that require a combination of critical thinking, logic, algorithms, coding, and natural language understanding.

I mean...singularity? Yes? No? Maybe? I'm not sure. But it's definitely a very cool accomplishment.

Big Tech Fractures

I truly do not know enough about the economy to write about it. But I do believe that big tech companies will begin to differentiate themselves based on how much AI is in their product or business. I think it's a good thing for the evolution of AI and for the adoption as well.

You can imagine tools that would assist teachers, advertisers, creatives, and so much more. The models are what's driving these tools and the companies with the best models will be the ones that succeed.

Conclusion

Overall, I'm very excited about the possibilities. I think if we have the "AGI" scenario we'll see a transformation of society never before seen (in a good way.) I think projects like AlphaCode, AlphaFold, and ChatGPT are going to be extremely practical today and will birth the next generation of tools.

I'm going to do my best to embrace the tools that will make me more productive. If you've got any recommendations, let me know!

and Happy New Years!