## 10 Stupid AI or Stupid programmers ?

I tried something that I thought would be very easy with all these smart search engines that can recognize everything... I took a picture of an artist from a google image search. I selected a part of the image and cropped it. I saved it and did another search with the google image tool, which is designed to recognize an image. Result.... all wrong with only images similar for the posture of the man but not the original image. Of course, most people will say that the "computer is stupid". But I don't forget that there is a team of people behind this tool.

How are you able to recognize and "translate" the content of an image and not recognize a part of an image in your database? It seems that the simplest search has been forgotten or is not the first answer that the search engine tries to find. The first problem, of course, is to know what the purpose of this tool in Google image is. The rare times I have used this tool has been to find out if an image is an original or a copy of something or somewhere else. For example, many image thieves use cropping to make watermarks disappear. I have seen it in some picture contests. I've also seen AI-generated images, but you can't prove it with a tool like this. It's more about light and detail.

Here it's just a question of the algorithm, a choice of priority levels of what you want to find in an image. There's no choice for the person using the search engine to say whether they want something similar or whether they want to find the origin or the subject of the image. It's a good example of how difficult it is to define a tool that responds to the behavior of a mass of users. When I define a tool in my job, it's only for 4 or 5 people, and I know their skills and habits. I try to make it simple, logical, and add help if they encounter some errors, problems. But when it's for so many different people, cultures, habits and skills, I don't know how they define their goals. Sometimes I get the feeling that they're making more difficult things just to impress the media and shareholders. Simpler functions become secondary. 

If you have examples of this, don't hesitate to contact me. All my projects start with the main function, which is done for me, and then I try to simplify it for the other users, with automatism, calculations, more attractive or easier to use form. My last "tool" is trying to interpret data from an acquisition device, converted to ASCII text. My problem is to exclude the parts where there is noise or 50Hz sinus. I had no time to do anything complex at the end of the year and I did it with a manual correction of two variables (a zone in the data causes errors because the device doesn't react as expected). I don't call it AI, of course, but some people might see some kind of AI in it. At the moment I feel stupid with this problem because it depends on many parameters and this zone is not the same for every data package. I have a lot of possibilities to find an answer, but my problem is ... that it has to be used by someone other than me. I'm very user-centric now and I'm not sure many programmers are when I se
e some GUI or web services. User experience is a very new job and it's not just a question of colors or shapes, it's a question of culture and skills. With AI everywhere, I'm afraid that AI will be misused to advise on decisions that are very human.

I'm very bad at programming because I only love the algorithm, not the code. But in the case I mentioned at the beginning, it's not about the code, it's about ... It's about knowing what the purpose of the programme is going to be. Maybe it's not a question for a programmer, it's a question for a manager, a product manager or something like that. The programmers probably did a good job for other functions, but someone just forgot that there are simple functions that are still useful and shouldn't be at the bottom of the algorithm result. What's more, I was anonymized so as not to have the effect of the famous 'search bubble' created with my own data. Is it useful ? 

2Dɛ

=> mailto:icemanfr@sdf.org Comments by mail
or by a reply on your blog