If you would like to join and invest in the goodness movement started by Help Steps: bit.ly/helpsteps-fonangels (collaboration)
Tim Urban has compared world history to a 1000-page book in his new book.A book with 250 years of history written on every page.Then he compared the first 999 pages with the last page. For the first 999 pages, our population was always under 1 billion.We have passed 8 billion on the last page. Throughout this book, we have used our own feet or horses and camels to get somewhere.On the last page, we first invented steamships, trains, then airplanes, automobiles, and finally rockets and landed on the Moon.There was no such thing as electricity throughout these pages, let alone electric cars et cetera.We simply cooked over the fire and lived by the candlelight. So, along with electricity itself, they were all invented in the very last page.Now I want to take this a little further.If we turn this last page, into a 250-page book.Do you know what would be written on its last page? DALL-E, ChatGPT, Whisper, Point-E, ChatGPT Plus and GPT-4. These are just some of the technological developments related to artificial intelligence that have occurred since August last summer. The times we live in are very different times.Just last week at least one new development was announced every day and we need to be able to read them very well.The last word added to these lines is GPT-4. It requires us to underline it and make a more careful reading.Yes, while we were looking around in a state of confusion about what to do with ChatGPT, while we had just made 17 different experiments of artificial intelligence; GPT-4 was introduced on the 14th of the third month, which is Pi day. Within 24 hours, striking examples began to emerge. It looks at the napkin and turns it into a website, what else? We will not only see examples like these in this video. We'll also capture some of the finer details in its 99-page technical documentation. Because now a new page has been opened in artificial intelligence. OpenAI claims "GPT-4 has human-level performance. Is it perfect? No. But you're not perfect either." Well, well. OpenAI says a lot as if it is a marketing employee lately. it's not like it used to be anymore. As the name implies OpenAI, started as, open artificial intelligence, after all. It was a non-profit organization at one time and they were trying to prevent artificial intelligence from being harmful to commercial concerns. There was even Elon Musk among its founders, but he later left. Now he may be saying "I wish I hadn't left". Because he asks, “How a non-profit to which I donated $100 million somehow become a $30 B market cap for-profit." I think he knows the answer to this question very well. Because he, like Walter White, turned into someone else right before our eyes. So, can we find the answer to this question?I even asked to GPT-4. It just is not aware that its new version is out yet. It's still not that self-aware for now.But there is quite a bit of innovation in many aspects.These innovations are grouped in 3 main categories. It developed itself in Creativity, visuality and length. I've shown you examples of creativity before. We know that It can do a number of creative and technical writing tasks, such as composing songs, writing screenplays, or learning and copying someone's writing style. That's why I like to call it author intelligence, not artificial intelligence. They gave an example from the fairy tale Cinderella. They asked it to explain the plot of Cinderella. But with a condition: Each word will begin with a letter of the alphabet from A to Z, in consecutive order and without repeating.
It is not known why anyone would make such a request, but it still fulfills this request to the point. We were able to request things from ChatGPT and the GPT3.5 on which it was based, just by typing. This new version accepts images as input as well. The length of the text entries has also been increased, meanwhile, we can now make entries of 25,000 words. So you give it a book of about 100 pages and you can ask for anything from it. In addition to these innovations, there are also some improvements. Tests measuring accuracy in machine learning have always been done in English until now. Because these machines mostly fed on and learned from English sources. This time, they asked 14 thousand multiple-choice questions in 57 different categories in different languages. Turkish is Among the languages with the highest accuracy rate 80%. The most successful language is still English with an accuracy rate of 85.5%, but we see that It has developed Its skills very rapidly in other languages, including Turkish.