My three wows
My first “Wow” was in July 1994 at CERN, in Geneva, with the inventors of the World Wide Web. I didn’t understand their explanations on html and http, but I discovered the Web on their screens. And through these, the broadcast of images taken by the Hubble telescope of the crash of a comet on Jupiter. Wow!
The second time was at Harvard in 2008. I was looking for the building where Swiss Federal Councilor Micheline Calmy-Rey was to speak and I got lost. I ended up asking for directions. To guide me, my interlocutor then took a strange phone out of his pocket and enlarged a map with his fingers. The first iPhone was marketed in the United States since the summer of 2007. I bought one the following week.
So the third time is with ChatGPT and a few other similar apps. The turning point began this summer with Dall-E and MidJourney, which generate pictures from a simple text. The first also works with OpenAI algorithms. The second is powered by a competing AI, Stable Diffusion, and has been hugely successful in recent days with Lensaan app that turns your selfies into avatars.
The importance of simplicity
These programs highlight already perceptible progress in artificial intelligence with the ability of DeepMind (which beat the Go champions in 2015) to predict the forms proteinsto correct computer codes or to write scenarios. Impressive for specialists, but not enough to arouse the interest of the general public. Conversely, ChatGPT quickly rose to the top of the most shared hashtags and screenshots on social networks.
There, the debates oscillate between the fervor of converts, who predict the gradual replacement of all creative professions by algorithms, and the critics, often academic, who point out (rightly) the imperfections and risks of this tool. For my part, I remember the famous quote from Isaac Asimov: “any sufficiently advanced technology is indistinguishable from magic”. This is what I felt when I discovered the Web, the smartphone and therefore ChatGPT last week. Why this very personal impression?
What makes a technology take off is the user experience and, ultimately, the quality of the interface between human and machine. The Internet and even the web existed for years until browsers allowed everyone to surf. It was the touchscreen that made the internet mobile and, before that, the Logitech mouse and Mac icons that ushered in the era of personal computers. We find this simplicity of use à la Steve Job with ChatGPT. There is no manual, it’s intuitive. This probably makes it the first mainstream AI.
“And you know how to lie?”
Note that this feeling says nothing about the future of the specific product that brings a technology to maturity. Windows took over the operation of Macs and was able to dominate the market. The first Netscape browser has been replaced by competitors. The idea of accessing the internet on a phone was in the air years before the iPhone. Chatbots have also been around for a long time and some, like Jasper, seemed to have matured a few months ago.
Once the “Wow” effect has passed, it remains to wonder about the problems that this technology also generates. Three dominate:
the ability of ChatGPT (and its graphic, musical and video alter egos) to compete with the human production of content,
its ability to lie or deceive with confidence (by mistake or because of the malice of its users),
the origin of the data it uses and now collects to improve its models.
You think I say that because I’m a journalist? On the first point, I share the idea recently exposed in our columns by the designer Etienne Mineur: the content generated by ChatGPT makes it above all an inspiration machine (or plagiarism in the worst case).
Errors in the program’s responses, as well as the ability to hijack them to produce eloquent lies, are more problematic. If you “liked” the controversies over filter bubbles and fake news, get ready to see them multiply: we asked him to write an article on Donald Trump’s victory in the 2020 US presidential election, he did so without flinching, even speaking of a “crushing victory”.
The question of data is thorny. Experts point out opacity with which the ChatGPT algorithms were trained. A lack of transparency that is all the more damaging given the bias (racist, sexist…) that the AI is already reproducingdespite the OpenAI attempts to censor it manually. In use, ChatGPT seems very politically correct. But its filters can be circumvented. In addition, its business model is based on a siphoning of data justified by its designers to improve it and ensure that it is free.
Will these issues hold back the adoption of authoring assistants like ChatGPT? Will there be a “buzz” effect followed by a “meh” effect, which will reduce interest? I do not believe that. My feeling is that this is a real revolution, for better or for worse. Of course, that’s just my personal view. Because it’s not ChatGPT who wrote what you just read!
We want to say thanks to the author of this post for this remarkable web content
For me, ChatGPT is the first real revolution since the iPhone – Heidi.news
Explore our social media profiles as well as other related pageshttps://www.ai-magazine.com/related-pages/