Google Gemini just lately bumped into bother for 2 causes: first, the AI chatbot returned biased solutions about Prime Minister Narendra Modi and second, it generated traditionally incorrect pictures. This led to huge criticism on-line, prompting the corporate CEO Sundar Pichai to convene a gathering through which he informed workers that the outcomes by Gemini have been unacceptable.It seems like these incidents have additionally prompted rival corporations, like Fb guardian Meta, to swing into motion.
In accordance with a report by The Data, with Llama 3, Meta is working to beat an issue it had within the Llama 2 AI mannequin.
“As a part of its work on the forthcoming model of its massive language mannequin, Llama 3, Meta is attempting to beat an issue perceived in Llama 2: Its solutions to something in any respect contentious aren’t useful,” the report stated.
Llama 2 safeguards made is ‘protected’
The report additionally highlighted that the safeguards that Meta added to the Llama 2 AI mannequin – which was launched in July final yr and powers the AI assistant in its apps – stop the LLM from answering a broad vary of questions which might be deemed controversial.
“These guardrails have made Llama 2 seem too ‘protected’ within the eyes of Meta’s senior management, in addition to amongst some researchers who labored on the mannequin itself,” the report stated, citing folks working on the firm.
Google’s AI ‘issues’
Google’s Gemini AI chatbot returned a ‘biased’ response to a query on Modi. A person on Twitter claimed that the AI chatbot didn’t present a transparent reply when an identical query was requested about former US president Donald Trump and Ukraine president Volodymyr Zelenskyy.
Google shortly swung into motion and stated that the corporate’s chatbot “could not all the time be dependable” in responding to sure prompts associated to present occasions and political subjects. It added that the groups at Google are working to refine the fashions.
Equally, the pictures created by the Gemini chatbot generated inaccurate historic pictures of the US Founding Fathers and Nazi-era German troopers. The pictures didn’t have any ‘White’ individual. Google introduced that it has paused the chatbot’s means to generate pictures of individuals and can launch a further-tuned mannequin.
In accordance with a report by The Data, with Llama 3, Meta is working to beat an issue it had within the Llama 2 AI mannequin.
“As a part of its work on the forthcoming model of its massive language mannequin, Llama 3, Meta is attempting to beat an issue perceived in Llama 2: Its solutions to something in any respect contentious aren’t useful,” the report stated.
Llama 2 safeguards made is ‘protected’
The report additionally highlighted that the safeguards that Meta added to the Llama 2 AI mannequin – which was launched in July final yr and powers the AI assistant in its apps – stop the LLM from answering a broad vary of questions which might be deemed controversial.
“These guardrails have made Llama 2 seem too ‘protected’ within the eyes of Meta’s senior management, in addition to amongst some researchers who labored on the mannequin itself,” the report stated, citing folks working on the firm.
Google’s AI ‘issues’
Google’s Gemini AI chatbot returned a ‘biased’ response to a query on Modi. A person on Twitter claimed that the AI chatbot didn’t present a transparent reply when an identical query was requested about former US president Donald Trump and Ukraine president Volodymyr Zelenskyy.
Google shortly swung into motion and stated that the corporate’s chatbot “could not all the time be dependable” in responding to sure prompts associated to present occasions and political subjects. It added that the groups at Google are working to refine the fashions.
Equally, the pictures created by the Gemini chatbot generated inaccurate historic pictures of the US Founding Fathers and Nazi-era German troopers. The pictures didn’t have any ‘White’ individual. Google introduced that it has paused the chatbot’s means to generate pictures of individuals and can launch a further-tuned mannequin.
