HOT NEWS/ Smuggled stones hidden in cocaine containers are discovered!

 Kliko këtu për të tjera
Voxpress:

The head of OpenAI, Mira Murati, speaks to the Italian media: I learned Italian from the television, like all Albanians

2023-04-05 10:45:33Lifestyle SHKRUAR NGA REDAKSIA VOX
The head of OpenAI, Mira Murati, speaks to the Italian media: I learned Italian
Mira Murati, 34 years old engineer, was born in Albania

 Mira Murati, the Albanian leading the development of technology that is changing the world, answers in Italian.

"My parents live in Italy and I learned it on TV, like all Albanians."

In short, ChatGpt artificial intelligence has been banned from touching the data of Italians.

"It was a surprise, they hadn't contacted us before," the 34-year-old engineer said through a Skype connection from San Francisco.

"We respect European laws on personal data, but we are open to dialogue and we hope to soon be able to restore our service in Italy."

The government claims that you do not inform users about the collection of their data and that you have no legal basis to feed them algorithms. What answer do you have?

We believe that our services are GDPR compliant and we actively work to reduce the collection of personal data in ChatGpt training. It is important that these systems learn the concepts we know It is a new and evolving technology, we work every day to make it safer and more accurate. Linguistic models are developed to predict the next most likely word, and in some cases it may not be more accurate, but the goal is precisely to get feedback from users, to understand inaccuracies to increase accuracy. We try to be as transparent as possible about this too: there is a message in the ChatGpt interface that makes the boundaries clear. When we launched Gpt-4 (latest version) three weeks ago, we let everyone know what the risks and potentials are to be

On March 20, there was a break-in to your systems that exposed users' personal data and even some of their credit card numbers. Were the Italians involved?

It was an open source component bug and when we found out, we immediately disabled ChatGpt to fix the problem. The bug also exposed the payment information of 1% of ChatGpt subscribers over a period of nine hours, I don't know if it happened and how much happened in Italy. But the full credit card numbers weren't showing anyway. The issue is now resolved and there is no ongoing risk to user data."

What are the conditions to reactivate the service in Italy? Are you willing to change anything in your systems?

We are working with the government to try to resolve this situation as soon as possible, we hope to have good news soon. But before the decision we had no dialogue, so it was a surprise for us. I repeat: we respect European rules, but we are open to finding solutions that address concerns.

Who does this blockade cause more problems: OpenAI or Italy?

From what I have read I understand that public opinion in Italy is in favor of ChatGpt. We do not want to leave Italy and hope to resume service soon.

Other authorities, from Ireland to Germany, are evaluating similar moves. Are you afraid of a regulatory backlash, in Europe or the rest of the world?

We are in constant contact with legislators and governments, in Europe and around the world, to understand how the technology should evolve from a regulatory point of view. I can't say what will happen in other countries, but I hope that if some governments have concerns they will be open to dialogue instead of shutting us down.

A letter signed by many intellectuals and tech entrepreneurs, including Elon Musk, asks you to stop training artificial intelligence for at least six months, for fear that the technology will get out of the control of the creators themselves. Will you stop?

We share many of the fears, having discussed the promises and dangers of artificial intelligence with various groups and government bodies for years. We are also concerned about the risk of an acceleration of technology. But there are two things about it: The first is that after we finished training ChatGpt4, we spent more than 6 months to make it secure before releasing it; the second is that now we are not training Gpt 5, contrary to what the paper says.

Many fear that ChatGpt will have impacts that could threaten the stability of our democracies, for example by eliminating millions of jobs. How do you handle this aspect?

Artificial intelligence will transform the fabric of reality and our society. Today it is mainly used as a support, as an assistant that improves one's work. But like other general technologies I expect it to change the nature of some jobs: perhaps it will create new ones that we cannot imagine, others will change. I don't think it's possible for a single company to solve the problem, but we take it seriously and do research to study possible interventions.

Another big danger is that AI gives anyone the ability to produce large amounts of "fake" content, which turns out to be disinformation.

It's one of the risks, so it's important to consult society early in the process and not develop these systems in the lab, in a vacuum, and then one day release them. It's important not to shock society with big changes in capabilities, but take people with you along the way: gather feedback, iterate and change along the way, expand access while gaining more certainty about safe uses. This is the strategy we follow: continuous and iterative development.

We have rarely been able to control the development of a technology: the desire for more powerful tools or economic interests are too strong. Should decisions be left to private companies?

"No, it's really important that governments and society as a whole are part of the conversation and the decisions. Technology is a tool, we can use it well or badly. It brings the possibility of something great and wonderful and also risks The solution is not to stop innovation: it is to understand it, to be able to anticipate its capabilities and to be responsible for its development.


Në përputhje me standarded JTI

Video

SHIKO KANALIN NE YOUTUBE
×