GenAI Trust: Failing Across All Trust Domains
Hallucinations are just one of many problems.
GenAI has a trust problem. With GenAI already contributing to fake news and deepfake videos becoming a source of revenue for pornographers, it’s hard to see how your bank’s new GenAI help system will get much love.
This isn’t an attack on GenAI. I love the tech, but it needs serious work before it gains trust, and GenAI hype seems to forget that conveniently.
Despite GenAI’s trust problem, I agree with Deloitte’s comment that “for many organizations, the risks of not embracing Generative AI outweigh the risks the technology creates.”
Companies should embrace GenAI as long as they plan to make AIs that can pass Deloitte’s six trust domains!
Let’s see how AI is doing in these trust domains so far by scoring them!
👉TAKEAWAYS
Fairness and impartiality: Score 0
“Limiting bias in AI outputs is a priority for all models”
We are certainly failing this test when Google’s Genesis model comes off the rails forcing CEO Pichai to say: “I know that some of its responses have offended our users and shown bias.”
Transparent and explainable: Score 2
“Generative AI-derived material or data, transparency and explainability also hinge on whether the output or decisions are marked as having been created by AI.”
Here, too, we are failing when nation-states, scammers, and pornographers are using AI without explicit notification that their content is AI-generated.
Score two points for China for requiring watermarks on GenAI content and that a movement is underway to make this the norm in the West.
Safe and secure: Score 4
GenAI presents the potential for inaccurate, misleading, or even harmful Generative AI-created content to be disseminated at scale.
Showing how high the stakes are, in France, a GPT-3 chatbot used on a medical line encouraged self-harm to a patient with suicidal thoughts.
Accountable: Score 4
Because an AI model cannot be meaningfully held accountable for its outputs, accountability is squarely a human domain.
If the AI isn’t accountable, who is? In the case (see below) X engineer Chris Bakke did not win his $1.00 car even if the chatbot said it would sell it.
Responsible: Score 0
Just because we can use Generative AI for a given application does not always mean we should.
Tell this to the fake news industry. On my news feeds for CBDC news, I get at least 2-3 articles daily that are GenAI-written, and fake. Let’s face it: So far, we’ve done a lousy job of policing GenAI use.
Privacy: Score 2
The data used to train and test Generative AI models may contain sensitive or personally identifiable information that needs to be obscured and protected.
GenAI IP theft is legendary, and while Game of Thrones creator George R.R. Martin's most recent lawsuit was dismissed, mine will follow! When I asked Microsoft’s “Copilot” for specifics about my book “Innovation Lab Excellence” it listed them freely. Hey that’s my IP unless Copilot bought the book!
X engineer Chris Bakke exploited a ChatGPT chatbot to get it to sell him a new car for $1.00. The car was not sold!
👊STRAIGHT TALK👊
I am not anti-GenAI, but I am against the never-ending hype that claims our transition to a new AI-enabled world will be easy and fast.
It will be neither. And the average score of 2 across domains shows how bad it is.
I have faith that the trust issues above can all be resolved to an average score of 7 within a few years, but is that fast enough?
Companies believe that “standing still means falling behind” and are launching bots as we speak. There are no audits or controls other than “take our word for it.”
The result is that we will all be at the mercy of flawed customer-facing bots until one or two fail spectacularly. Only lawsuits will force change
Trust is fragile. Once a few of these bots come off the rails, you’ll start to find push—back and calls for reform. But only after trust is broken
How long will it take for the machines to win it back?
Thoughts.
Copilot apologizes for stealing from my book Innovation Lab Excellence almost word for word! It listed all twelve of my Innovation Lab Best Practices!
Thank you to all my subscribers who have been sharing Cashless!
You are now the No. 1 driver of subscriptions, and prove that word of mouth still works better than X!
Join the community by subscribing! You’ll be glad you did!
As with many transformative technologies, standing still means falling behind.