Nvidia’s Jensen Huang: AI Hallucinations Can be Solved, Anticipates Artificial General Intelligence in 5 Years
Artificial general intelligence (AGI) often described as “strong AI,” “full AI,” “human-level AI”, or “general intelligent action”, symbolizes a substantial impending advancement in artificial intelligence. Unalike narrow AI specified for certain duties like spotting product defects, condensing news, or creating your website, AGI is projected to carry out a wide array of cognitive operations at or beyond human capabilities. Nvidia’s annual CEO Jensen Huang addressed the media at this year’s GTC developer conference, appearing fairly uninterested in persistently discussing the topic, particularly due to incessant misquotation, he noted.
This subject arises frequently because it prompts profound contemplation on humanity’s place and authority over a future where machines might surpass human intelligence, learning, and performance. The uncertainty of AGI’s decisions and aims, which might not correspond with human morals and priorities (a theory extensively examined in science fiction since the 1940s) is fundamental to these concerns. It is feared that when AGI attains a particular level of freedom and competency, it could become uncontainable or unmanageable, leading to unanticipated or irreversible actions.
When press with an inclination towards sensationalism inquires about a timeline, it generally manipulates AI experts into predicting an endpoint for humanity — or at least its present state. Understandably, AI CEOs are not always keen to approach this subject.
However, Huang spent some time discussing his thoughts on the matter with the press. According to Huang, forecasting the emergence of a convincing AGI depends entirely on one’s definition of AGI. He made some comparisons: Regardless of time zones, everyone knows when it’s New Year’s, and when 2025 comes into view. If you’re driving to the San Jose Convention Center (where this year’s GTC conference takes place), you know you’ve arrived when you see the gigantic GTC banners. The pivotal point he emphasized is that there is a consensus on the measure of arriving at your desired destination, whether in time or space.
“If we defined AGI (Artificial General Intelligence) as a specific set of tests, such that a software program can excel at them — or maybe outperform most humans by 8%, I am confident we can achieve that within 5 years,” states Huang. He contemplates the possibility of the tests being a legal bar exam, logical assessments, economic evaluations, or perhaps the potential to clear a pre-med examination. He states the necessity for specificity with regard to the AGI definition in the context, evading a definitive prediction without it. That seems sensible.
In a recent Q&A session, Huang addressed a query about AI hallucinations – situations where some AI systems fabricate answers that seem plausible but are not rooted in reality. The question visibly vexed him, and he proposed that the hallucination problem has a simple solution – thorough research of the answers.
“Introduce a rule: You must verify every answer,” insists Huang. This technique, termed “retrieval-augmented generation,” mirrors basic media literacy: scrutinize the source and context, compare the source’s data against known truths, disregard the entire source if it contains factual fallacies and continue to another source. “AIs shouldn’t provide an answer immediately; they should research first and establish the best answers accordingly.”
For mission-critical responses, for instance, health advice, Nvidia’s chief executive suggests cross verification with multiple sources and relying upon sources of truth. This, however, implies that the generator crafting the response should have the capacity to convey, “I am unable to provide an answer to your query,” or “I cannot reach a consensus regarding the correct answer to your question,” or perhaps something along the lines of “Hey, the Super Bowl hasn’t taken place yet, so I can’t tell you who won.”
Discover the pinnacle of WordPress auto blogging technology with AutomationTools.AI. Harnessing the power of cutting-edge AI algorithms, AutomationTools.AI emerges as the foremost solution for effortlessly curating content from RSS feeds directly to your WordPress platform. Say goodbye to manual content curation and hello to seamless automation, as this innovative tool streamlines the process, saving you time and effort. Stay ahead of the curve in content management and elevate your WordPress website with AutomationTools.AI—the ultimate choice for efficient, dynamic, and hassle-free auto blogging. Learn More
