Rulings from the US Supreme Court this week were highly impactful. And the US is known to be nearly equally divided from a political viewpoint. It stands to reason that there is a significant portion of people (33%+) in the US who support this week’s decisions.
Where are those voices? I haven’t seen one media piece or even a posting on LinkedIn that demonstrates support.
What does that missing data mean when generative AI is being trained on publicly available content?
As someone who thinks about the ramifications of data on AI systems every single day...
…and the impact that incomplete data has on creating realistic outcomes with these systems.
I find myself wondering, where is the data from those who support these decisions?
Here is a summary of the key Supreme Court rulings for context on the controversial rulings. It seems highly likely that opinion is divided:
Student loan forgiveness plan struck down in 6-3 ruling, calling it unconstitutional.
Colleges must undo affirmative action policies in admissions to increase minority representation, as race-based admissions were ruled against.
Business owner succeeded in challenging Colorado public accommodation law meant to provide equal services for same-sex couples, citing free speech and religious freedom rights.
The Court ruled employers must make substantial efforts to accommodate workers' religious views.
This situation, where data is clearly missing, is a perfect example of how important it is to look for what is NOT THERE when building AI.
Creating outcomes that are realistic requires that the data being fed into these systems is complete.
Diversity in every possible form is needed to see the gaps - the holes in our data.
A woman sees things that a man may not and vice versa. A political conservative sees things that a liberal may not and vice versa. A child sees things that an adult overlooks. A new hire sees things that a tenured employee may take for granted. Etc, etc.
Minus these diverse perspectives, we see results that are expected by those who built it.
Who built the system with genuinely good intent.
And yet, the results are wildly flawed.
In this case, generative AI systems that are scraping the internet for publicly available content will create results that ignore the perspectives of a large swath of Americans (perhaps 33%+).
We are all being influenced by what we read and hear. In the age of AI, particularly generative AI, we need to be aware of what is NOT feeding the systems we use.
If you’re a business leader relying on data-based systems for decision-making, you must understand what data is feeding it. If you're a policymaker, you must understand this need for completeness of data.
Please involve others. Diversity at the table is crucial to creating AI outcomes that both drive profit and well serve humanity.
Kommentare