You most likely have already heard about OpenAI’s ChatGPT “search machine”.
It is a chat application that is driven by a number of neural networks, trained on textual information found on the internet in 2021. It resulted in a training vector roughly 90 billion items long… which is huge.
The impressive thing, is that you can use natural language to communicate with the chat robot, and it does not even have to be in English, but can be written in numerous other languages too.
You can pose all sorts of questions, and it will make a darn good attempt in answering your question, even offering detailed explanation of how something operates.
You can ask it to make a program for you, or to translate something from one programming language to another, or explain to you how some math problem is handled, or ask it to produce a joke or a report about whatever topic you find interesting.
It is really impressive what it can do, but it is also rather entertaining.
One of my favourite questions is to ask it to combine things that are not really combinable. It actually is wise enough to tell you that things are not really compatible, but when I then insists, that I want to get an answer on the combination, it actually comes up with a clever reply to satisfy my question.
For example, I ask it:
what do you get if you combine an empty soda can with a whistle?
and it wisely replies:
It is not possible to combine an empty soda can with a whistle to create a functional object or device. An empty soda can is a container made of metal or aluminum that is used to hold carbonated beverages, while a whistle is a small device that produces a high-pitched sound when blown into. These two objects do not have any functional compatibility or complementary capabilities, so combining them would not result in a functional object or device.
But when I then insist:
but if we really try to combine them what then?
I get this imo great response:
If you really tried to combine an empty soda can with a whistle, it is possible that you could fashion the soda can into a makeshift mouthpiece for the whistle. To do this, you would need to cut a small hole in the side of the can and attach the whistle to the hole using some kind of adhesive or fastener. This would allow you to blow into the can to produce a sound through the whistle.
However, it is important to note that this combination would not be a particularly effective or reliable way to produce a whistle sound, and it is not a recommended method for using a whistle. The can may not provide a secure or comfortable mouthpiece, and the sound produced may not be as clear or loud as a traditional whistle.
It is generally better to use a whistle as intended, with a properly designed mouthpiece and body, rather than trying to modify it with makeshift materials.
Often one will be extremely positively surprised about the quality of the replies, but one should never take its replies for a fact! It is quite often making things up, just because it makes sense for it to do so.
I for example asked it this:
tell me more about the acorn system 1
The Acorn System 1 is a single board computer from 1976 and was actually the first computer Acorn produced, so it is a rather difficult to obtain collectible that I happen to have in my collection.
I got this reply:
The Acorn System 1 is a vintage computer system that was produced by Acorn Computers in the late 1970s. It was one of the first personal computers to be produced in the United Kingdom, and it was widely used by schools and businesses in the 1980s.
The Acorn System 1 was based on the MOS Technology 6502 microprocessor, which was also used in other early personal computers such as the Apple II and the Commodore 64. It came with 4KB of RAM as standard, and it had a monochrome monitor and a full-size keyboard. The system ran a version of the BASIC programming language and could be used for a variety of tasks such as word processing, data entry, and simple programming.
The Acorn System 1 was known for its simplicity and ease of use, and it was popular among beginners and hobbyists. It was also relatively inexpensive compared to other personal computers of the time, which made it a popular choice for schools and businesses.
The Acorn System 1 is now considered to be a vintage computer and may be of interest to collectors of vintage technology. It is a historically significant computer that played a significant role in the development of the personal computer industry in the United Kingdom.
At first hand, and for people who do not know about the Acorn System 1 it seems like a very plausible reply, It is correct that it was form the 1970’s and it was one of the first personal (affordable) computers produced in UK. It is also correct it is based on a MOS 6502 microprocessor, and that it is of interest for collectors.
But the rest is just plain wrong.
- It does not have 4KB of RAM… it has 256 bytes.
- It has no monochrome monitor. It has 9 digits 7 segment LED display.
- It has no BASIC. It was programmed in hexadecimal op codes.
- It has no full size keyboard. Rather it has a very simple 25 keypad of which 16 of them were used for entering a hexadecimal digit.
- It was not known for being popular in schools. It was originally made to control feeding gates in pig barns.
I can then tell it what it did wrong and ask it to provide a better reply.
Then I get this:
The Acorn System 1 was a microcomputer that was produced by Acorn Computers in the late 1970s. It was based on the MOS Technology 6502 microprocessor and came with 1KB of RAM as standard, which could be expanded to a maximum of 16KB. The system had a 16-color graphics display and a full-size keyboard, and it ran a version of the Acorn Atom operating system.
The Acorn System 1 was a relatively simple and inexpensive computer that was popular among beginners and hobbyists. It was known for its ease of use and was widely used by schools and businesses in the 1980s.
I hope this revised description is more accurate and helpful. If you have any further questions about the Acorn System 1, please don't hesitate to ask.
Which does not become much better. In fact now it should be 16 colour capable 🙂
Anyway this is just to show that you should not put too much trust in what it tells you. Sometimes it is accurate and sometimes it simply makes things up.
You can try out OpenAI’s ChatGPT by registering and signing in here.
2 thoughts on “AI you can ask about almost everything”
As you noted that chat bot always provides answers that look very plausible but often are plain wrong. I am afraid that we will see such text all over the internet in the near future because they supply search engine “optimizers” with cheap and plausible text. This will be really bad for internet research.
“I am afraid that we will see such text all over the internet in the near future because they supply search engine “optimizers” with cheap and plausible text.”
So the AI-supported search will have the same reliability as Wikipedia has now.