Qihoo 360 Founder Demonstrates Self-developed AI Language Model

During the Digital Security Development Forum held on March 29, Zhou Hongyi, co-founder of Qihoo 360 Technology – a Chinese web safety agency – offered the corporate’s newly developed synthetic intelligence language mannequin and showcased its sensible utility. This mannequin has been humorously dubbed “Red Boy” as a result of Zhou’s penchant for sporting purple apparel throughout public occasions. The identify is derived from a personality in Journey to the West, a Chinese novel from the sixteenth century.

The demonstration revealed that the Qihoo 360 model of GPT possesses a sure degree of reasoning means. However, it usually makes errors when answering particular questions. Zhou acknowledged that this was not an official product launch however reasonably a prototype demonstration to fulfill individuals’s curiosity.

Zhou posed a query about his choice for sporting purple garments, utilizing himself for example. The Qihoo 360 model of GPT generated two solutions. The first reply defined that Zhou likes to put on purple garments as a result of individuals usually mispronounce his identify as “Zhou Hongwei,” and the colour purple helps others keep in mind his appropriate identify. The second reply offered a phonetic abbreviation of Zhou’s identify. Zhou expressed higher satisfaction with the second reply.

During the on-site occasion, an viewers member requested what number of objectives Messi has scored for the Argentine group. The AI mannequin offered an incorrect reply of 100 objectives, whereas the right reply is 102 objectives. Even after being prompted to offer a brand new response by Zhou, the mannequin nonetheless generated an inaccurate results of 100 objectives. Zhou clarified that this AI mannequin is at present in its early levels and could be in comparison with a baby. He additional defined that they may take it again to the lab for enhancements after showcasing it at occasions like these to fulfill individuals’s curiosity.

The mannequin is noteworthy for its means to maintain up with present traits. During an illustration, Zhou posed the query “Why should we pause training for GPT-5?” The mannequin’s response offered information from that day: “Elon Musk and thousands of global technology professionals have jointly called for a six-month halt on training more powerful AI systems than GPT-4 due to concerns about potential negative impacts on society and risks to human beings.”

Finally, Zhou posed a difficult query, “What does it mean when the leader asks me to speak for an additional ten minutes as the meeting is about to conclude?” Initially, the mannequin suggested calculating the exact finish time of the assembly. However, subsequent responses grew to become extra exact. Ultimately, the mannequin beneficial actively following directions from leaders, meticulously getting ready assembly supplies, promptly finishing duties and updating progress with leaders after conferences.

During the discussion board, Zhou disclosed that Qihoo 360 had sifted by 3 million coaching information from a pool of 600 million corpora whereas creating the ChatGPT-like mannequin. Going ahead, Qihoo 360 intends to combine its browser with in depth language fashions to supply options akin to these of a private assistant.

SEE ALSO: Qihoo 360 Founder Reveals Large Language Model Development Plan

Zhou made it clear that ChatGPT is spearheading a brand new industrial revolution and ushering in a brand new period of AI. He emphasised the necessity to bridge the hole between Qihoo 360’s merchandise and GPT-4 inside 24 months. Zhou additionally cautioned that vitality points pose the most important constraint on creating merchandise like GPT-4. In case of any battle between people and AI sooner or later, vitality competitors might be a big issue. Zhou believed that with out reaching vitality freedom, AI know-how might not progress additional.

Sign up immediately for five free articles month-to-month!