ChatGPT Over a Year Later

Do you use ChatGPT or other, similar tool? I can say, I use it almost every day and for multiple purposes. Not only because of work – but it is probably the most beneficial in this area, but also in my personal life. To check some information, to get advice, to prepare trip plans or even to not forget about something important, when I will prepare to something. During last year it was transformed from some nice, maybe a bit exotic stuff into powerful tool we can – and maybe even should – utilize to simplify our lives and be more efficient. I know there is a lot of articles about ChatGPT, but I’ve decided to write what I think about that, also with some comparison to other “revolutions” we observed in last few years.

In autumn 2022 ChatGPT has been released, and a lot of changes started. In last few years we had a lot of big stuff and new technologies, but not all we’re really interesting. For example, Web 3 and block chain. Most of startups or projects related to them completely died, a lot of initiatives were total scam (like NFTs) and in reality, they do not resolve any important problems. Almost everything provided by Web 3 is possible also in standard solutions like “classic” databases and in such cases, this is just reinventing the wheel. All of it in times, when people talk more and more about ecology, and we have cryptocurrency which requires a lot of energy (for CU, GPU, memory or storage based on solution). Finally, quite huge entry threshold, because it was required to configure some wallet in the browser, so it is far from user-friendly solution for masses. I was never really interested and after some time I think I made a good choice. Of course, it always a risk to ignore something new in It, but we also cannot focus on everything new, it is even not possible.

Buzzword or real change?

With AI tools and assistants, it was different – it is because they really change a lot of stuff, can help us during daily work and improve a lot of processes. Even children use them to resolve their homework Then mobile apps have been added to use it more easily and finally everything has been connected. Right now, we do not need to select we want to generate image, browse internet, analyze files or just reply some questions. ChatGPT do this for us under the hood, by analyzing questions. On mobile we can just use voice to ask questions and it’s much better than previously available assistants like Google, Siri or Alexa. I know they can do in other languages, but in polish usage of for example Google Assistant was in the most of cases asking about current weather… So, nothing really big and valuable.

AI-based solution is like having REAL personal assistant, coworker. It does not need to generate completely new demand (like tablets) and then try to convince us. It just helps in our existing problems, and it changes everything. Some people quickly stopped using tools like ChatGPT after initial trying because they think, they do not provide right answers. This one is a bit complex – they can do a lot, but prompts we use are crucial element. If we will write super simple questions, without details, without discussion, answers will be the same. Do you see similarity to real life conversations? It works in the same way: if we will give more data, it can be analyzed and we can provide better options, better possible decisions.

Tools like ChatGPT have also some “control techniques”. For example, there is a big difference when we will write “prepare example codebase for… App” and “working step by step, prepare example for… App”. It is because such models work great in incremental steps, but we need to inform them, they should use this method. I effect answer will contain information about thinking steps and we will see better why some decisions have been made. Also, thanks to next questions we can make adjustments, work together to fix some issues and in effect, have really good work. It is not a magic wand solution, or maybe sometimes it is, but we still need to know how to manipulate such wand using our hands. I think these “limitations” can be removed in coming, close future. It will be next big step, but it will be still neural network with a lot of unpredictable stuff inside, similar to human brain. As a father, I can cay unpredictable is totally normal if we want to act with thinking being.

Drawbacks, threats

Of course, I also see some drawback, and even threats. First of all, some people can lose their jobs – editors, SEO content writers, some junior developers, maybe even some layers. What I can recommend is the book “Average Is Over” by Tyler Cowen. I think it was visionary, because mentioned about coming changes, even if only providing info about chess algorithms and AI usage in this area. There is interesting scenario proposed by this book: person alone will not be enough… AI will not be able still to do all stuff… But mix of them can by something amazing, also in job market. What about future? It is hard to say, but I can imagine even scenarios like in Terminator movie. AI can learn much faster than humans, so it will not need us. And remember about our history, we have tendency to fight each other, and such threats should be eliminated.

Second item is we need data to learn models. So far companies provided them a lot of research publications, internet stuff and of course, our activity. But what if we will decide to stop creating, because everything can be done by AI? It can still work, but models must be extended to use more abstraction, to work similar to human brain, and be very creative. In other way, everything will be very similar, will constantly reuse the same ideas. It is visible even now with few image generators. After observing some of them, I can easily say which one has been generated by AI because they always have their specific form, not similar to real photos or many other creative creators.

There are also more and more interesting models provided by different companies. Some of them we can use locally without any payments, of course not in such comfortable way, because they provide just API, without any graphical interface, but for engineers it is not a problem. Thanks to this API, they can build completely new application which can utilize even some specialized models to help us resolve specific issues. Even something like AI assistance on phone photos editor – I remember over 10 years ago I was in one of conferences when Adobe presented Content Aware function from Photoshop, option like magic eraser, to remove something from photo. It was simple, not always perfect but if someone was familiar with Photoshop, was able to use it and improve a lot of photos. Right now, similar option is on multiple devices, much faster, much simpler and big software like Photoshop is not required for that. It is just simple example, how it can work.

Even Big Techs can miss this

AI “revolution” showed also some interesting thing – some companies, even with a lot of required data can be behind without any gold reason. I think great example of that is Google. Their Bard service was… Much worse than ChatGPT and there were some funny situations almost immediately after release, even if it was much later than ChatGPT 4. This company has tons of data because advertising and their personalization is their fundamental service. They know how to collect it, operate on it, and they also have tons of users which just provide this info by using emails, search, maps, YouTube, Chrome browser and many other services.

They already had assistant app for a few years, they already worked with machine learning, and it was visible in Translate or reCAPTCHA, but in this “race”, they missed something big. Why? I do not know, but situations like with Kodak are still possible – if you are not familiar, I will just notice it was the biggest analog cameras manufacturer with total domination on the market. When digital cameras arrived, they decided to ignore them, because it is only temporary, not important trend and people will still prefer analog… Bug mistake and in effect, Kodak almost stopped exist. There are many other scenarios like that from the last 20-30 years, but it always was the same: someone had domination so did not react properly on new features, new trends.

Is AI “must have”?

In my opinion tools like ChatGPT is not just a nice feature but even must have for developers and many other positions. It can improve our work a lot, so we can work more efficiency or just do more and easier. Of we will not use them, other people will still do, and we will completely behind. I think not technical skills, bit mix of soft and hard skills is the most important but, in this situation, it does not matter, because such tools can help us always. I’m also not a fan of rat race, if I will see something like that, I prefer to say goodbye because I know it is not worth it at all – I have only one life and do not want to waste it only for job and couching talks. But again, it can help also in personal life a lot, so it is a real game changer.

I started with ChatGPT as an example and want to finish also with that. Right now, Plus version costs $20 per month and based on how much it can bring, I think this is very, very low cost. If you work on IT industry or something which can be assisted by some AI tools and think it’s too much, you make a huge mistake in my opinion.