Jonathan Gillham
speaker
95 appearances
1 recordings
1 series
first heard Nov 2024
last heard Nov 2024
Jonathan Gillham’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsNo recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.
Appearances
The their political view, if you took all the GERD rails off, you know, their political view would be the average of the content that had been consumed.
And so it's it does create some really interesting questions.
And we did we have a an ongoing study where we look at what websites are blocking the different AI bots and
the AI bots will drift in the direction of their training data in terms of their their biases and and political views.
And what we saw was that the majority of sort of the center and left uh media were were are were blocking uh GPD bot and the majority of sort of the right leaning meet news outlets were not were not blocking.
And so that raises sort of an interesting question on would that would that over time, if that were to continue, would that bias the direction at which the bots were leaning?
And then would people put other other Gerdrails up anyway?
So it's as a really interesting it's a it is a really interesting question right now that we have in the world around if we will all sort of use these um you know, LLMs as a as a brain for how we create and consume and
and make opinions on things, the direction that those will move, will that move society in that in that same direction.
You know, I think we weren't concerned about social media being capable of doing that when it first came out, um, you know, two thousand eight.
Did we think Facebook would be capable of sort of influencing elections?
And I think people would have sort of laughed at that.
Um, but we, you know, we clearly saw that it did have that level of power over society.
And that we should be concerned about it.
And I think people are concerned and aware of the same thing that if if every time that we use AI to produce content, if it has just a slight bias in a certain direction, will that then be reflected in the content at a large scale that gets published and will then eventually move move society in the direction that was either intentionally architected into the LLM or just uh a
function of the the the data that the LM was trained on.
Yeah, so we we've been we refresh it, refresh it daily.
So we look at sort of AI bot blocking.
Um and I'll just pull up the the stats, but some what we've seen is that yeah, it was sort of it quickly jumped up to about 15%, capped out.
March of this year at around 35% of the top 1,000 websites were blocking GPT bought.
Showing 41–60 of 95 · page 3 of 5
← Previous
Next →