Listen to the audio version of the article
Mira Murati is a technical profile, an expert in artificial intelligence and will do well. Everyone is convinced of this, both experts and markets. Sam Altma’s replacement at the helm of ChatGpt is 34 years old, she is a mechanical engineer with previous experience at Goldman Sachs and Tesla and a year ago she was chief technology officer of Open Ai so she knows her chatbot well. Although she was announced as an interim – or temporary – appointment for the transitional phase, she is the right profile to put OpenAi back on track. Although as The Verge writes, some investors would like the return of Sam Altman as CEO, probably also to defuse the risk of a departure of employees and managers if the father of ChatGpr decides to launch a new startup. We don’t know exactly the reasons that led to the dismissal of its co-founder but we can imagine them.
The latest signs were the blocking of new accounts on Gpt-4, the paid version of ChatGpt and a month of disruptions that have continued to the present day. Sam Altman’s latest releases had already hinted that something wasn’t working. Defined by the New York Times as the Openheimer of our times for having invented a device (generative AI) potentially capable of destroying humanity, as has been feared on several occasions by more than one protagonist of AI Gen, Sam Altman he began to be more and more cryptic and visionary. Someone will remember when he defined the hallucinations of the AI, i.e. the incorrect answers of the chatbots, not as errors but as characteristics of the software. “Hallucination are not bug but feature”, he said which is a bit like saying, it’s not them (the Ai) who make mistakes, it’s us humans who don’t get it. However, the breaking point probably occurred with the release of Gpts, customized versions of ChatGpt. Think of an App Store model where, however, the individual App can be “trained” by the customer with their own data. In fact, you can upload up to 10 documents (max 512 Mb each) with useful information for the AI to carry out a specific task. A few days after the release of Gpt Platform, some users began to report some problems. For example, users who use a GPT are able to trace the information and prompts that its creator used to generate his AI. It means exposing the competition to its strategic assets. And then in the OpenAI terms of service there is no reference to prohibitions regarding copyrighted materials. And then there is a cost problem. As you know, we don’t know how much the energy impact of Gn Ai. Those who are using GPT to power other services have experienced slowdowns and inefficiencies. The impression is that the choice to enter the market early with a “generalist” chatbot is showing the negative effects that competitors such as Google had suggested. The information is not segmented and continues to arrive from a user base that has quickly exceeded 100 million. More questions means more computing power. As some experts have warned online at the moment, cht gpt is like an uneducated child that everyone wants but who still doesn’t have the basics to run and risks being uncontrollable. This is perhaps why someone talks about how the greatest technological innovation of last fifty years risks becoming the largest marketing operation of the century.
The new CEO of OpenAI who, we remember, contributed to transforming the company from a startup to one of the most important artificial intelligence companies in the world is called to bring order, to set boundaries for ChatGpt. Maybe it’s a bit like when Steve Jobs was removed from Apple because he was too visionary. Years later, after he returned, he said it was the right choice.