With the recent interest in generative AI, we haven’t heard much about Microsoft’s contribution as one of the Big Four companies. Although Microsoft released Cosmos for processing text and images with generative intelligence, it chose open artificial intelligence applications as its partner in desktop applications, which is its most popular product ever.
Microsoft’s choice seems to come from complete conviction in the capabilities of open artificial intelligence applications, the most important of which is the application of Chat (GPT), as evidenced by a recent report issued by researchers at Microsoft, in which they conducted an accurate evaluation of the capabilities of Chat (GPT), from which they came out with an exciting conclusion. The title of the report sums it up: The Glamor of Artificial General Intelligence.
The researchers reach an important conclusion after many tests, that the application of Chat (GBT) is an important step towards general artificial intelligence, and in the definition of artificial intelligence, access to general capabilities is a long-term goal, as it reaches the possession of general capabilities that are similar to human capabilities or exceed it. , which is a dangerous and destructive ambition in the opinion of some.
The report lists a number of tests that were conducted for Chat (GPT) in various fields of mathematics, drawing, and interaction with the world. Some of the tests are serious that end in wit and funny that end in novelty. Perhaps what the report reached, as described by the team leader, is the suspicion that leads to Astonishment tinged with frustration, which may end in fear, so you think: oh my God, where does all this come from?
In one funny example, the report lists how the application helps the user solve the problem of water leakage from the kitchen ceiling. The application answers: Check if there is a toilet on the upper floor or another source of water. After the user answers that the kitchen is located under the toilet: Check the toilet for signs of leakage in the floor such as bubbles, drips from the ceiling, or stains on the walls. And the application continues to give instructions with questions until the user reaches the solution.
The researchers test aspects of the drawing, for example, so they ask the application to draw a shape, and it returns what is required, and if the same drawing is given incomplete, and the request is repeated again, then it completes the drawing as the first time, indicating that it also sees, what amazes the researchers is the impression it leaves on the user that it Talking to a customer service employee, is this impression justified, or does the app easily trick us with some stereotyped phrases?
The team goes far in testing the application, as it was asked to conduct a Shakespearean dialogue between two people about proving that the prime numbers are endless, and if the answer came in full, the researcher was no longer astonished to be sure of who he was talking to? It gets confusing as someone says: something unknown is doing something we don’t know what it is.