Was there a recent big advancement in AI?
We asked our Michał Łusiak 10 questions on the state of AI and ML. Why the sudden public interest (ChatGPT anyone)? Has AI really advanced that much in the last few years? Where is all this going? What should organisations do about it? Will this affect jobs? What are the opportunities?
1. What is, in your opinion, the state of AI right now? Was there a recent big advancement? How? If not, why does it feel like it was?
I wouldn't say there was a recent big advancement, especially if one has followed this area. The fact is that most people see AI as "computers doing funny things". However, because tech generally accelerates in bursts, and the bust of ChatGPT into people's lives, it made AI more tangible and accessible.
AI is generally a marketing term, and I feel it's a bit of a white rabbit - we're chasing it, but we can never catch it. When the first computers showed up in the 1940s (for calculating naval artillery fire), people were saying "it's not AI, it's just simple math". When Gary Kasparov lost in chess against Deep Blue in the 1990s, people were saying "it's just an algorithm". Rise of deep learning that allowed some cool things such as image recognition and finding patterns, was labeled as "just statistics". Now the general public slowly got aware of how ChatGPT is working and they say "it's not AI, it's just a large language model". As humans, we don't want to admit that something is intelligent, but the truth is that things are improving - a lot. Machine Learning (ML), this is a more technical term, experienced steady growth since probably 2010 when deep neural networks proved to be useful - thanks to growing amounts of data and cheaper and cheaper computing power. There are new algorithms popping up every week doing better use of underlying statistics to do something that might be perceived as "magical".
2. How are things compared to 3 years ago?
I think this is a great time frame to discuss. Because around 2020 only a few data scientists realized that Transformers (a type of ML algorithm, originally invented at Google in 2017) are not only good for natural language processing (NLP), but can also be used for graphical processing and few other things. Those applications were traditionally separated in ML and used with completely different approaches, and now we suddenly found a unifying "theory". I personally remember seeing it as a big moment, but didn't realize where this was going. This gave the foundation for those algorithms that generate art/images from prompts, and describe images into text. We connected those two worlds exactly in the last 3-5 years. But it wasn't explosive. There were many papers, a lot of work, some funny results - until they stopped being funny.
3. What do chatGPT and all those new hot AI startups mean for us as people, society or organisations?
I think, like with all the tools, there will be good things that will come out of it and there will unfortunately be misuse too. Companies should learn how to use the tools to automate processes and become more efficient. The effects on our society are harder to predict. For sure people will lose jobs, like it happens with every technical innovation. But new jobs will be created as well. For example, some companies already talk about "prompt engineers" - people who are skilled in writing prompts to AI algorithms to get good results out of them. Generally we're in constant acceleration of tech and it's becoming harder and harder for our minds to catch up and this is what will be challenging.
Peter Diamandis famously said: “There will be two kinds of companies at the end of this decade... Those that are fully utilizing AI, and those that are out of business.”
4. Are companies in general more aware of the power/possibilities of using AI than before?
I feel that many companies have very little idea how to use even simpler forms of "AI", that we have had around for around 20 years. We're in a bit of a hype bubble in the tech industry and we feel that everybody's using it, but even in tech the awareness of what's possible seems pretty low. That's why we get exaggerated positive reactions such as the one to ChatGPT - in the end it's only a chat interface to tech that's been around for a year.
5. What are some very cool examples that you have come across recently of AI use in companies to either optimise or make revenue or everything in between?
https://interiorai.com/ - post a photo of your room, get AI to redecorate it for you in chosen style - for now you pay for renders, but in the future they could monetize ordering the stuff you need for that decor. It's using the custom mod of the stable diffusion algorithm used to generate art/images from text. Very close to home usage that most people understand, fun to use, cheaper than protein-based intelligence.
https://avatarai.me/ - upload your photos, generate images of yourself in different settings. They were doing it for a few weeks, then Lensa came with their iPhone app and took the market with its price. If you are looking for higher quality, you should give it a try.
Github co-pilot is interesting for "our line of work", it helps you write code from prompts. For example "I want to sort that list" and it generates the code. Doesn't solve bigger architectural problems in programming, but makes mundane tasks much, much easier and quicker. It uses very similar concepts to ChatGPT, but it is trained on code (btw, ChatGPT can write code too).
6. Where do you think this is going? What is exciting about it? What are you the most excited to see happening?
I think what's interesting for me is that this will boost the value of professions with "human touch". Psychotherapist (although some people try to use ChatGPT for help), general "care" (healthcare, elderly care), etc.
There will be more and more things automated (store checkouts, etc), which will gradually reduce daily human interactions. We're built to have those interactions and we'll seek them. This gives a chance to people who are good at, well, people.
All around medical tech is cool too. AI is getting better than doctors at reading X-rays and similar. AI is also helping develop drugs for rare diseases by creating custom proteins. And much more.
7. What is the biggest challenge now?
Wrapping your head around it. There's a lot happening. Lots of solid stuff, but also many hyped empty promises. Finding things that are useful and practically applying them is a challenge. Long term maintenance of those tools too. Lots of small threads like this that we have not gotten our heads around yet.
8. What are the possibilities that non-digitised companies should think of right now? Is it now or never when it comes to doing something about it?
I don't think it's now or never, but getting digital gives you higher chances of survival, to be more efficient and to be profitable. I'd say the sooner, the better. When it comes to possibilities, find things that use most of your time and resources, find ways to automate them. Rinse and repeat.
9. Is AI for every company? If not, what types of companies or industries benefit the most?
Companies that already collect data, or have their parts of businesses digitized will of course benefit the most - because this gives them a faster and more solid start. If you still keep everything on paper, that transformation will of course be more challenging.
10. Anything you want to add?
When it comes to copyright of AI generated content, where do we stand on that matter? This would be an interesting next article that we should make.