Pichai: AI images 'unacceptable'

Google CEO Sundar Pichai sent a memo calling the company's AI image mistakes "completely unacceptable" and said it plans to launch new processes for AI launches. Pichai called the issues “problematic” and said they “have offended our users and shown bias." Pichai’s memo said the teams have been working around the clock to address the issues and that the company will instate a clear set of actions and structural changes, as well as “improved launch processes.” You can find the full text of the memo here

Google’s artificial-intelligence push is turning into a reputational headache. The online backlash around its chatbot shows why the company hesitated to release the technology.

Google Gemini debacle proves my point... A processor cannot understand the meaning of a single word, let alone the meaning of any sentence it builds. To a processor every word is an array of characters represented internally by numbers, the processor is doing the only thing it can i.e. arithmetics and comparing numerical arrays when it statistically regurgitates nonsense from its data set, if the data set and the statistical model are in any way flawed. This approach is fundamentally flawed and can spread misinformation if relied on. Why are AI engineers hell-bent on getting the computer, which can process only quantities, to deal with natural human language which is not at all quantitative? AI is best used in engineering applications where the processor's quantitative capabilities and speed serve as an extension to human intelligence. I am not saying that AI is useless, I only think it shouldn't be used for natural language. The whole exercise is grotesque, to put it mildly.

Googles chat bot Gemini did poorly on philosophical questions . Current AI models are clearly not ready for prime time. They need to get better to avoid entrenching bias. This was a flaw too with big data

In my opinion, it will be hard to find an AI model good for everything , just as humans they need to specialize in something to bring better outcomes…for marketing purposes , defining your value proposal with for who and how is a simple strategy…and it better won’t be vague , so that your weaknesses won’t make headlines ….

Google’s artificial-intelligence push is turning into a reputational headache. The online backlash around its chatbot shows why the company hesitated to release the technology.

Even Google (who created the attention transformer architecture that powers GenAI) is not immune to the many risks that come with deploying GenAI products and services. Google wanted to take a more measured approach with GenAI deployment, but they got pressured into moving quicker with the success of OpenAI's release of ChatGPT. Before deploying GenAI products and services, your company should reflect and determine what level of risk your company is comfortable taking on with respect to GenAI. Many are comfortable with significant risk. Many will want a more measured approach. Neither is right or wrong - but your company should determine what is best for it and use that determination as a north star for all GenAI decisions.

Google is in the news for an embarrassing misfire of their Gemini generative AI model. They temporarily paused its ability to generate people after accusations that its guardrails were mistuned. However, this raises an inteinteresting question: what if we just agreed you shouldn’t be able to use generative AI to generate people? One of the first viral generative AI images featured the Pope in a puffy jacket. Others have featured Presidents Biden and Trump hanging out as best buds (I think they are probably not). In other instances, people have used these tools to generate nonconsensual explicit imagery, the most famous being the ones about Taylor Swift. While there are some great use cases like “generate an image of five consultants around a white board” or “show me a soldier in a futuristic command center” or “take a photograph of an astronaut riding a horse,” you could actually prevent a lot of misuse if you said you can generate artistic scenes and flowers and pictures of poodles dressed as Iron Man. But what if you said no Iron Man?

With technology shaping our everyday lives, the recent controversy surrounding Google's AI tool, Gemini, has ignited a crucial conversation on Diversity, Equity, and Inclusion (DEI) in artificial intelligence. While striving for inclusivity, we must also navigate the complexities of historical accuracy and representation. It's a delicate balance, but one that is essential for creating technology that reflects the diversity of its global user base. Let's use this moment to reflect on how we can better integrate DEI principles into AI development, ensuring that our technological advances are inclusive and representative of all. #artificialintelligence #ai #DEIinTech #AIInclusivity #TechnologyForAll #hitl