You know that fun trend where you share those old, quirky photos of yourself back in the greatest year on earth, 2016? Yeah, well, you may be inadvertently training AI models with them. Because we can never have anything nice.
Show More Show Less View Video Transcript
0:00
Your 2016 throwback pics? They're cute
0:02
And they might be training AI. So you've likely seen people posting pictures of themselves from 2016
0:07
and we see this every year at some point. Nostalgia sweeps the internet, people post photos of themselves when they're younger
0:13
It happens all the time. This year, though, when users were once again encouraged to submit photos of their 2016 self
0:18
the trend faced some backlash. Experts accused it of providing AI research companies
0:22
with free, high-quality data used to train models on aging, facial recognition, and style trends
0:28
And it's true. These kinds of posts can hand over more than a decade of facial aging data to algorithms
0:33
It helps AI companies because it solves one of AI's biggest challenges, temporal data
0:38
And you're doing all the work for them. Oftentimes, labeling images and videos for AI training is expensive and time-consuming
0:45
but it's made way more efficient and affordable when people dump their images online and label them so clearly, like with this trend
0:52
Users are voluntarily providing photos that are matched pairs with identity confirmation in their profiles
0:57
and then providing additional data, like facial geometry and lighting conditions and backgrounds and even metadata
1:04
There a decade of real biological aging in these photos not just cosmetic change This kind of data helps AI models better see how people age and better recognize identities over time So the more data points an AI has
1:17
the better it works as a predictive model, like matching old photos of people to new surveillance footage, for example
1:23
This might feel unimportant and honestly, maybe you're fine with your old photos training AI data sets
1:27
But as Nicholas Papernot, an associate professor of computer engineering and science at the University of Toronto told Times Colonist
1:33
Something that feels benign to you right now might not feel benign to you forever, since
1:38
quote, we don't necessarily have a perfect way of anticipating what technology will be
1:42
available and what will be the applications of that technology. Now, don't blame yourself for participating in the trend
1:48
Some studies suggest that people become more nostalgic when they go through hard times, and we are experiencing nothing, if not hard times
1:55
In fact, nostalgia has been associated with an increased meaning in life, self-esteem
1:59
and optimism. But it is true. The nostalgia constantly sweeping the internet could be a pull for your data, and there are
2:06
conspiracies that AI firms started this trend on purpose, but I find that pretty unlikely
2:12
We see this trend every year, so my guess is that it probably began organically and AI is just
2:17
taking advantage of the trend. One thing we all love posting is ourselves
#Celebrities & Entertainment News


