“Kejutan Manusia dengan Komentar Tak Terduga tentang AI senilai Rp 2.000 Triliun”

Jakarta, CNBC Indonesia – Nvidia is the company that benefits the most from the development of artificial intelligence (AI) technology. The American chip giant is a supplier of AI chips to tech giants such as OpenAI, Microsoft, Meta, and Google.

In its performance reports over the past few quarters, Nvidia has consistently broken all-time records. This has propelled Nvidia CEO Jensen Huang’s wealth to US$123.8 billion and made him the 9th richest person in the world according to Forbes Real-Time Billionaires.

On various occasions, Jensen has extolled the benefits of AI for the future of humanity. This includes when he visited Indonesia and spoke at the Indonesia AI Day 2024 event.

However, recently Jensen made an unexpected comment about AI. He said that the AI currently available cannot provide the best answers to the information in the world.

“Currently, the answers we get are not the best versions that AI can provide,” Jensen said in an interview at the Hong Kong University of Science & Technology, as quoted by Yahoo Tech on Monday (25/11/2024).

He stated that the world still needs a few more years for AI to be truly reliable.

According to Jensen, people should not doubt the answers provided by AI. Currently, many question whether the AI answers are hallucinations or not, whether they are appropriate or not.

“We need to reach a point where the answers you get can be fully trusted. I think it will take a few more years to reach that point. Currently, we will continue to improve our computing,” he explained.

Large language models (LLMs) like ChatGPT have significantly improved in recent years in their ability to answer complex questions. However, there are still limitations in their functionality.

For example, AI chatbots often provide answers that are hallucinatory or contain misleading information. This is a major issue with AI that is being continuously addressed by its developers.

OpenAI, generally regarded as a leader in the AI era, was sued by a radio host last year because ChatGPT created a false lawsuit about him.

Other AI companies are also facing lawsuits due to errors in their LLM systems. The current limitation of LLMs lies in processing and training large amounts of diverse data.

In an interview last week, Jensen mentioned that pre-training or training models on large and diverse datasets before developing them to perform specific tasks is essential.

“Pre-training that simply takes all available data and automatically extracts knowledge from it is not enough,” he said.