What's up with this whole AI thing?

February 18, 2024

This page is targeted at a non-technical audiance.

Right now, AI is moving crazy fast in multiple areas:

• Generating images in most any style, including photo-realistic to the point of deception
  • This is quickly moving to include video generation as well
  • This also includes manipulating existing images
• Generating descriptions of photos
• Generating speech from text in a way that is almost totally convincing
  • This includes making new speech models from existing recordings of voices
• Generating near-human quality text responses to textual input
  • Chat bots, obviously, - but a whole lot more.
• Generating transcripts from audio - speech to text

I want you to take a second to stop and consider some of the implications of that.

The bad things - political fakes, fraud, etc - is probably what comes to mind first, but also think of some of the good.

You may have realized these advancements are more than the sum of their parts, we can hook them together.

There are already systems where you can send a picture of your math homework which includes a diagram necessary to understand the problem and have the system understand the input and give you the correct answer in text, along with steps to get there. (Actually, we’ve had these for some time, but they’re now getting scary good)

*BUT, the answer may be wrong and the steps may be bullshit. Depending on the complexity, it’s likely to get it right, but it may not and it will sound confident when incorrect.

It’s possible to make a computer where you talk to it and draw it pictures and give it text, and it does the same back.

That’s not sci-fi anymore.

But I think that still doesn’t quite put into perspective - it’s still too abstract and you still probably have this pre-conception about what it means to talk to a chat-bot.

This isn’t your typical chat bot. It’s a chat bot that can do many very difficult, lots-of-schooling needed tasks in a few seconds. This includes giving good medical advice, financial advice given data, and writing code. That last point, I think, is particularly interesting even if you never intend to write code.

If you’ve spent any significant amount of time using a computer, at some point you’ve wanted to some menial task involving a boat load of files and just said “Nope, fuck that”. For those of that know how to code, we might write something to do it for us, but out of laziness and value for our time we’re also likely to say “Nope, fuck that” too. But we don’t have to value an AI’s time:

But I am 1000% confident soon even this will be overly complicated. You’ll just be able to select the files, hit the “AI” button, and say “Rename these files to be the title” and it will just work, because it will write code like this on the fly and run it for you. It’s also likely it will be able to try multiple times, checking it’s own work if it doesn’t get it right the first time.

That flow where you ask for something and it writes code to do it, tries it, checks its own work, repeat if necessary is likely to be totally transparent to the user soon. This means you’ll be able to ask the computer to do things that if you couldn’t program you simply wouldn’t have known how to do before.

I doubt it’ll be making a full video game for you any time soon, but it can definitely assist you in every step along the way of making one.

Of course, it can do pretty much anything else too:

• It can write beautiful poetry or give you feedback on your own.
• It can watch as you type in a spreadsheet and suggest formulas
• It can read a whole page of text and help you skip to the part you care about
  • Yes mom, that means you can skip to the recipe without reading someone’s life story.

Remember early smart phones? #

When the first iPhone came out it was obvious that it was the future, but it took everyone some time to actually make it good. App developers had to solve new problems that arose because using big meaty fingers to touch a screen was a lot different than using a full size keyboard and mouse. Web pages had to be redesigned to handle the vertical layout and smaller interface. This took time to figure out.

That where this is at right now. We have the tech - the models exist and there’s no doubt they’re the future - but they’re not integrated into society yet, at least not well. That’s the billion trillion dollar problem that Microsoft, Apple, and Google are all desperately throwing money at to be the first past the post, because once we have something with this tech fully integrated, not using it would be like going back to dumb-phone.

We’re currently at the launch of the iPhone first gen. There have been other smart phones - BlackBerry, etc. - but this is the inflection point. It’s still not in everybody’s hands, some people haven’t caught on (and some never will) but it’s a defining moment. You have the pre-smartphone and the post-smartphone era. This is that again.

Tech Bros #

As you’ve read this, you may be thinking “I’ve heard this all before, BitCoin was the future too”

GOD. FUCKING. DAMNIT.

All of those cryptocurrency morons share the same handful of brain cells. I’m actually not sure that after the end of covid that didn’t all just wind up somewhere drinking koolaid chanting their praises to the almighty Elon.

I do not say “This is the future” lightly. I also, and this is vital to understand, didn’t say it’s a future I want. I think it’s really cool and can do some real good and it can also do some harm. I’m not invested in it’s success in a literal, monetary sense. I’m only invested in my friends and family not falling behind, so I’m trying to help them see what’s coming so they can keep up.

I am also already finding it greatly helpful for a small but not insignificant number of tasks and want to encourage others to at least try it.

Good or Bad? #

Bad:

1. It uses an ungodly amount of power causing insane carbon emissions
2. It can stop people from making genuine, human connections
   • Yes, there are already “AI Girlfriends” and that situation is weirdly complicated
3. People blindly trust the output
4. Kids may over-use it in school and not learn
   • This could cause more problems as people don’t know how to check if the output is good, making a feedback loop with the above point.
5. There are criminal uses, including
   • Making AI-nudes of people without their consent
   • Faking voices of loved ones for scams
6. It is already shot-blasting the internet with low-quality ad-farm shit making it harder to find good, human authored content
7. currently, the good ones can only run on what is damn-near a super computer, meaning it’s running in ☁️ the Cloud™ ☁️, so you can’t really control it and it may require a subscription fee.
   • This, hopefully, won’t stay the case.

Good:

1. It can review human authored content and find potential mistakes.
2. It can be used artistically as a tool, not a replacement (inspiration, not generation)
3. It makes doing some tasks much, much, much faster. Especially menial tech work.
4. It’s a great teaching aid - it can explain things quite well and prompt further questions about understanding.
5. It can be used by the blind to ask what an object is
6. It can lower the barrier to entry for extremely technical positions and make difficult, technical work faster provided the user can adequately review the output
   • I have already technical perused projects I wouldn’t have before because they wouldn’t have justified themselves with their time-cost to benefit before

So, the big question, does the good outweigh the bad?

I think yes. I think it will finally allow normal people to take advantage of computer in a way that will reduce the number of people doing bullshit jobs (nobody in 2024 should have a job of “Data Entry” when a computer vision system is more accurate, faster, etc.), allow people in technical roles to make better products faster, it will help doctors treat patients, etc, etc.

I also think, or hope, that while some people suck, the percent of people that suck is massively outweighed by the percent of people that are awesome. AI is a tool, and any tool can be misused. So long as more good is coming from it than bad, well, why not, right?

How do I try it? #

The simplest, easiest to use tools you can try right now, for free, with no prior skill in using these are ChatGPT and DALL·E 2. Note that DALL·E 2 is starting to show it’s age: There are much better image generators now, but it’s easy to try.

Both of these models are locked down. You can’t ask them to generate anything not safe for work, at least not without getting your account banned.