When Elon Musk, you don’t have to rely on centuries of prevailing human understanding – you can create your own.
“We will use GROK 3.5 (we may call it 4), which has advanced thinking, to rewrite the entire human knowledge collection, add lost information and delete errors,” MUSK wrote at X Friday night.
After that, he said that he would re -train the last Grok model at the new knowledge base to be free of similar waste. He added, “A lot of garbage in any foundation model is trained in unrelated data,” he added.
MUSK has sought for years to create products such as Twitter and renoed Grok, which are free from what is seen as a harmful prevailing restriction.
Business Insider previously reported that the Grok Army from “AI Teachors” was training robots on a set of topics of discounts to compete with Chatgpt “woke up” from Openai. MUSK on Saturday’s request from X users to respond to its post with examples of “division -making facts” that can be used in Grok Re -Training.
Gary Marcus, a critic of the noise of artificial intelligence and an honorary professor at New York University, compared Musk’s efforts with the Orwellian tariff, the first time that it is compared.
“Directly from 1984. You cannot get Grok in line with your personal beliefs, so you will write the date to make it compatible with your opinions,” he wrote on X in response to Musk.
It can be to enjoy renewing traces in the real world.
In May, just as Musk was returning to his work in Washington, DC to re -concentrate his various companies, Reuters reported that Duji was planning to expand his use of GROK to analyze government data.
“They are asking questions, and they get it to prepare reports and give data analysis,” a source told Reuters, referring to how the robot is using. Two other sources told the port that officials at the Ministry of Internal Security were encouraged to use it despite the fact that it was not approved. He told the representative of the new republican administration that “Doug did not push any employees to use any specific tools or products.”
Grook also had security problems. In May, after the company said it was an “unauthorized amendment” to the back interface, the robot began to refer frequently to the “white genocide” in South Africa. The company soon replaced and said that it conducted a “comprehensive investigation” and was “the implementation of measures to enhance Grok’s transparency and reliability.”
Xai did not immediately respond to a request for comment from Business Insider.
