There is a joke among AI experts that OpenAI, the company behind ChatGPT with its most recent language model, GPT4, has stopped being “open”, open in English, to be “closed”, closed. For the first time since it was founded in 2015, it has not provided information on the magnitude of GPT4. This time they have not communicated anything about the number of parameters with which their AI works, nor with what data they trained it. Just a small mention that they use a supercomputer whose power is not known either.
The hypocrisy became capital in the messages of Sam Altman, CEO of OpenAI, when he said this week in an interview that he is concerned that his company will not be the only one that develops this technology and that there will be others “that do not put some of the limits of security” that they do put. So he warns that society “has a limited time to figure out how to react to it, how to regulate it, how to manage it.”
OpenAI asks that the sector be regulated – that is, its competition – while it is the first to launch new access modalities to its AI without waiting for any regulation, either with ChatGPT and Bing, the Microsoft search engine, or through Internet services. third-party companies that use it with their clients.
The list of problems that the lack of control over such a powerful technology will cause is enormous. A few days ago, someone proposed to GPT4 to give him a programming code so that the human could execute it and get a release from the AI ??and his internet connection. Guess the machine’s response: delighted with the proposal, it wrote the code for it. Imagine if you were asked for software that could be used for cyberterrorism.
Just yesterday, Eliot Higgins, founder of investigative journalism site Bellingcat, was banned from Midjourney, an image-generating AI, for asking the machine for photos of an alleged arrest of Donald Trump and his subsequent prison break. The journalist spread 50 of those false images on Twitter.
The posting of fake images and videos of real people, which look shockingly authentic, is one of the biggest dangers we are going to face with increasing frequency. But there are others.
OpenAI has produced a report on the professions that are endangered by their technology. In general, many that require high degrees of training. The same manager who today plans to replace a skilled worker with AI can also be replaced by it. The professions that the machine cannot yet replace are those that require manual skills. The Biblical curse “you will earn your bread by the sweat of your brow” is still very much in force.
