Bot or not?

Monkey!The last comment on this blog is from September last year. A certain Ricky posted a comment on my tide clock post. The strange thing is that the comment is on topic, but still misses the point. The last sentence is that the person would like to see the result. But ‘the result’ is exactly what the post was about.

What a creative and fun project! A tide clock is not only practical for beach lovers like you but also a unique piece of decor. It’s great that you’re taking the DIY approach to create something that perfectly suits your taste. The step-by-step process you’ve outlined seems both simple and rewarding. I can imagine it will be really satisfying to see the tides reflected on your custom-made clock! Looking forward to seeing how it turns out.

It looks like this comment is generated by some kind of AI, where this kind of comment has only one goal. To get on the list of accepted comments, so that in the future comments from this user are automatically approved. Once that is achieved, there will be a tsunami of messages about gambling, sex, crypto, and medicines arriving on every page.

Posted in AI

AI: Are we there yet?

A mind inside a fishbowl on top of a mechanical body walking through the sky on top of a cloudArtificial intelligence is a thing we keep hearing more and more about, but what are we really talking about? Is it really something smart and intelligent. It is certainly smart, but sadly the intelligent part still needs to be created.

A few years ago we had various very good search engine. You could give it a couple of words, and it would use those words to really quickly give a list of results that were related to the words you provided. They improved this search by remembering where people clicked. More clicks meant a result was probable better, so it got a bit higher score.

Now OpenAI has added a new layer to this. It’s tool ChatGPT can understand text, and can tell you something in a natural fashion. You can ask it a question, and it will give you an answer back, just like a normal conversation. Understanding someone is a very complex task. You can ask the same thing in lots of different ways, and small details can be very important or meaningless. The ‘engine’ is trained to know what is important, and what is not. Instead of giving you a list of sites, now you can simply get a summary as an answer. The AI is not storing a huge amount of links, it is reading everything, and storing the information it finds, which is what it uses to answer your question.

We are not there yet, but it is a huge step forward. Understanding language is really difficult, and making it seem ‘natural’ is very hard, but it looks like they really achieved something here. Still, there is no intelligence. For now there is no curiosity, no initiative, no desire. The AI only responds to queries. It remembers what you asked it, but thats it. You can ask it things, and it can tell you things, but it won’t do anything else. It is really good, and clearly the current ‘AI’ is much much better then the ‘old’ search engines.

We can even ask it directly.

Q: are you really intelligent?
A: As an AI language model, I don’t possess a conscious intelligence like humans do. However, I have been trained on vast amounts of data and algorithms to respond to various queries, so I can simulate an intelligent conversation with humans. I can understand natural language and can generate responses based on the information I have been trained on. My abilities are limited to what I have been trained on and what algorithms are available to me.
Q: can you repeat the answer in dutch
A: Als een AI-taalmodel …