AI stocks under pressure? NVIDIA CEO explains: Why trustworthy AI is still years away

• NVIDIA CEO warns of risks of artificial intelligence
• AI systems like ChatGPT not trustworthy
• According to Jensen Huang, reliable AI is several years away

Artificial intelligence has made enormous progress in recent years and is now present in many areas of investors’ daily lives. But despite this rapid development, AI is still far from being truly reliable. NVIDIA CEO Jensen Huang made the comments in an interview with the Hong Kong University of Science & Technology in late November 2024, stating that it will take “a few years” to develop an AI that we “broadly trust.” can.

A persistent problem of AI: Why ChatGPT hallucinates

Huang explained that today’s AI systems are not yet able to consistently provide the best and most reliable answers. “The answers we have today are not the best we can give,” said the NVIDIA CEO. This problem is particularly evident with large language models (LLMs) such as ChatGPT. Despite the impressive advances in natural language processing and the ability to answer complex questions, there remain weaknesses that limit the reliability of these systems.

A central problem with AI-controlled chatbots is the so-called “hallucination”, in which the models provide false or invented answers. A notable example is an incident where ChatGPT created a fictitious lawsuit about a radio host. The radio host then sued OpenAI. Such errors raise doubts about the reliability of AI systems and show the importance of continuing to improve these technologies.

NVIDIA CEO explains: That’s why investors will have to wait years for reliable AI

For Huang, it is crucial that AI systems one day become so precise and reliable that people no longer have to question whether an answer is correct or “hallucinated.” “We have to reach a point where you largely trust the answer you get,” Huang emphasized in an interview with the university. While the progress in AI development is undeniable, Huang says more years of research and improvement are needed to realize this goal.

Another aspect Huang highlighted is the need to further increase computing power to train AI models more effectively and expand their capabilities. Huang also emphasized that simply pre-training a model with huge data sets is not enough. “Pre-training – simply taking all the data in the world and automatically discovering knowledge from it – is not enough,” he said. Rather, it is necessary to further develop the AI ​​models in a targeted manner and not just rely on large amounts of data to achieve the desired results.

Trust in artificial intelligence – a distant goal?

Despite the rapid development in the field of artificial intelligence, there is still a lot of work to be done before we can experience truly trustworthy AI. The “hallucinations” and uncertainties that still frequently occur today make it clear how far we are from an AI that we can trust without hesitation. But with advancing research, improved computing power and thoughtful further development of AI models, this goal could be achieved in a few years.

Editorial team finanzen.net

This text is for informational purposes only and does not constitute an investment recommendation. finanzen.net GmbH excludes any claims for recourse.

Selected leverage products on NVIDIA

With knock-outs, speculative investors can participate disproportionately in price movements. Simply select the lever you want and we will show you suitable open-end products on NVIDIA

The leverage must be between 2 and 20

Advertising