TECH NEWS – The generative (content-creating) artificial intelligence chatbot formerly known as Bard has had its capabilities curtailed by its creator.

Recently, Gemini came under fire for starting to generate inaccurate, harmful images. So Google decided to resort to the old nerf tactic of dumbing down its image-generating capabilities until it could somehow fix the outstanding bugs. This is a big mistake for the company because Gemini is treated as a standalone product by Google, so it has no direct connection to the search engine, AI models, and other services the company offers. Google has manually fine-tuned the model to make image generation diversive, i.e. not biased towards people of certain groups or ethnicities. This would allow people from all over the world to use the technology without encountering such problems.

Only Google’s countermeasures could not prevent what happened, because Gemini was unable to generate images accurately (and without racism) when it received a query for a particular skin color. This resulted in odd queries (e.g., “Black teacher in a classroom”) that led to inaccurate, even offensive images being generated (or refused to work) because the model treated them as sensitive when they were not.

Google has therefore decided to disable the ability to generate images of people in Gemini. The company is now putting the entire model through extensive testing and will not restore the chatbot’s full functionality until Gemini no longer fails. The company has also said that the AI models are not perfect, and while they have improved a lot since they were introduced, they still tend to “hallucinate” and this can cause problems.

Google has not yet said when it will be ready to fix Gemini, so until then all we can do is wait. What we can hope for is that by the time Gemini is fully functional again, it will be in a much better state and have even stronger generative capabilities.

Source: WCCFTech, Google Blog