
요약
Although LLMs and similar systems have impressed us, they haven't yet demonstrated real humanlike smarts
본문
Deep learning has driven several key breakthroughs in AI, powering capabilities like image recognition, multilingual translation, and even real-time conversational partners.
But despite these impressive demonstrations, real intelligence still eludes our most advanced systems, which continue to excel in what's essentially an advanced form of curve fitting.
"The goal of strong AI is to produce machines with humanlike intelligence, able to converse with and guide humans. Deep learning has instead given us machines with truly impressive abilities but no intelligence. The difference is profound and lies in the absence of a model of reality."Fake it until you make it
The esteemed computer scientist Judea Pearl was reflecting on the original vision of intelligent machines with the AI systems that we have today in the first chapter of his book.
Latest Videos FromTechRadarWatch full video here: Quote of the dayThis article is part of TechRadar Pro's QOTD project to provide an insight into the minds of the brightest and most recognized figures in the technology industry today and in years gone by. Read the full series here.
Released just a year after Google scientists' seminal paper outlining the transformer architecture that underpins modern large language models (LLMs), 'The Book of Why' explores how causality changed science. Pearl was a leading AI pioneer and helped devise the probabilistic approach at the heart of deep learning – going on to receive the 2011 Turing Award for his contributions to the field.
His thoughts, on the whole, are that AI systems aren't anywhere close to true intelligence despite what comes across as outstanding mimicry, and that they're effectively limited due to a fundamental lack of understanding of cause and effect – instead relying on statistical patterns from which they generate outputs.
Dawn of superintelligence
Despite Pearl insisting that existing AI systems aren't anywhere close to reaching artificial general intelligence (AGI) or similarly advanced milestones, many AI execs have recently insisted that we've either reached that point or that we're close to doing so.
OpenAI president and co-founder Greg Brockman and ex-Nvidia CEO Jensen Huang both claimed we were at AGI with the launch of recent AI models. Then, we were on the cusp of impressive models like the GPT-6 Astra family and Google's Gemini 4 models. But they still adhered to the same frameworks and architectures that have powered similar LLMs.
Are you a pro? Subscribe to our newsletterUS President Donald Trump also declared that the term 'artificial intelligence' was redundant because we had now reached 'super intelligence' – but ASI is already defined as a concept in science, and there's no basis for this milestone being reached.
TOPICS
Keumars Afifi-Sabet is a freelance contributor for Tech Radar and the Technology Editor for Live Science. He has written for a variety of publications including ITPro, The Week Digital and ComputerActive. He has worked as a technology journalist for more than five years, having previously held the role of features editor with ITPro. In his previous role, he oversaw the commissioning and publishing of long form in areas including AI, cyber security, cloud computing and digital transformation.
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.
LATEST ARTICLES- 1 007 First Light composers share their stance on AI music in video games — ‘We feel art is only interesting if it is made by real people, informed by their experiences of life’
- 2 Quote of the day by computer scientist Judea Pearl: "Deep learning has instead given us machines with truly impressive abilities but no intelligence"
- 3 Tiny Corp's refurbished MI300X tinybox packs 768GB of HBM into a desktop box, and the price has already jumped by $29,000
- 4 Save $500 on Acer's ProDesigner PE320QXT 6K monitor for Black Friday
- 5 AMD is adding tamper-resistant hardware to its chips to protect from physical attacks