Recently, an artificial intelligence scholarAndrew Ng(Andrew Ng) In an interview with This Is The World, he wrote: AGIIT HAS BEEN OVER-PRODUCED AND REDUCED TO MARKETING TERMS, AND IT IS ASSERTED THAT THE INDUSTRY WILL NOT BE ABLE TO ACHIEVE REAL AGI IN 2026。
Recently, in an interview with The Is The World, the artificial intelligentist Andrew Ng pointed out that AGI (General Artificial Intelligence) has become too much a term for marketing and asserted that no real AGI can be achieved in 2026。
DURING THE INTERVIEW, U NANDA CRITICIZED SOME OF THE CURRENT BUSINESSES FOR FREQUENT MANIPULATION AND LOWERING OF AGI STANDARDS FOR PUBLIC RELATIONS OR FINANCING PURPOSES。
HE PROPOSED A NEW "TURING AGI TEST": IF AI CAN PERFORM ECONOMICALLY VALUABLE TASKS INDEPENDENTLY, LIKE SKILLED HUMAN TELEWORKERS, IN THE COURSE OF MANY DAYS OF EXPERIENCE, IT WILL MEET THE SOCIAL PUBLIC'S LEGITIMATE EXPECTATIONS OF AGI。
U Nda stated that it was decades away from achieving that goal and that, as open Internet data approached drying up, the path of increasing intelligence simply by increasing the size of the parameters was being challenged。
In response to subsequent technological evolution, U Nda points out that the core commercial values for 2026 and beyond will be concentrated on "smart workstreams"。
In his view, it would have a more definite economic benefit than the blind pursuit of an all-powerful single-body model, by giving the existing large language model tools the ability to call on and guard the vertical areas of legal compliance, medical assistance and customer services in a step-by-step manner。