Wow, Fred. I was concerned before. But now you have me terrified. After doing a long tour in Silicon Valley where side-effects of programming aren't really considered and the urge to move fast and break things is strong, I worry that for all the good ways this technology could be used, its bad uses are far more likely to be practiced. And legally again, we will be found asleep at the wheel. There is a right to privacy in the Constitution but corporately it doesn't seem to mean much any more so I fear we will continue to thoughtlessly slide down this slippery slope.
I stubbornly cling to the notion that if we know in advance about new technologies we can work to shape them in positive ways.
There is much to overcome, as you point out Sandra. But profit isn’t everything and an informed society is still capable of resistance, especially on behalf of the generations to follow us.
The growing pushback on addictive social media, on AI in the arts, on surveillance, on pornographic deepfakes, etc, is heartening.
It’s extraordinary that with all the media available so little is known about so many important issues. Let’s keep on trying to inform the public and push for better outcomes.
Thanks Fred, great article! The progress in this field has indeed been interesting. Over the past 20 years I think the main progress has been due to progress of AI models to analyze (and encode/represent the data as well as de-noise/correct the results). The normal challenges seem to remain in these studies to date regarding size and types of training and test data sets, unknown reproducibility/performance in real world settings (not to mention the cost and contraints of required fMRI/MEG machines), even though the newer LLMs and AI models can help ignore reconstruction errors. But with more money being thrown at these problems, and emerging interest for BCI in large profitable corporations, and the investigation of newer types of neuroscience-focused AI foundation models, the future will be interesting. Regarding the privacy issue, these systems record what the person is consciously imagining, at least in "normal" subjects, so perhaps less of a concern at least so far is my thinking, unless you accept the "memory download" potential some have been advocating. Without my rambling on too much, the most valuable initial use cases appear either for accessibility for physically challenged people, or new use cases we are yet to think. For example, for most people it is easier, cheaper, and far more reliable to use your hand to control a switch rather than to use brain signals for the foreseeable future, the advantage, should reliability improve is for applications requiring ultra-low latency (e.g. professional gaming), but reliability cannot be compromised. Sometimes truly good use cases for some new technologies can take decades to develop in the absence of substantial marketing budgets :)
I agree with all that you write. But it is difficult to gauge the speed of change in this digital era. I am still amazed that AI-generated photorealistic imagery has become commonplace just three decades or so after the emergence of Photoshop. Now that seems relatively slow.
Dear Fred, Thank you for this piece, a lot to chew on. I have wanted a camera to be implanted in my forehead long ago when I realized I saw the photograph through my eyes not a viewfinder, and was lucky if the image stood still long enough for me to record it.
Wow, Fred. I was concerned before. But now you have me terrified. After doing a long tour in Silicon Valley where side-effects of programming aren't really considered and the urge to move fast and break things is strong, I worry that for all the good ways this technology could be used, its bad uses are far more likely to be practiced. And legally again, we will be found asleep at the wheel. There is a right to privacy in the Constitution but corporately it doesn't seem to mean much any more so I fear we will continue to thoughtlessly slide down this slippery slope.
I stubbornly cling to the notion that if we know in advance about new technologies we can work to shape them in positive ways.
There is much to overcome, as you point out Sandra. But profit isn’t everything and an informed society is still capable of resistance, especially on behalf of the generations to follow us.
The growing pushback on addictive social media, on AI in the arts, on surveillance, on pornographic deepfakes, etc, is heartening.
It’s extraordinary that with all the media available so little is known about so many important issues. Let’s keep on trying to inform the public and push for better outcomes.
Thanks Fred, great article! The progress in this field has indeed been interesting. Over the past 20 years I think the main progress has been due to progress of AI models to analyze (and encode/represent the data as well as de-noise/correct the results). The normal challenges seem to remain in these studies to date regarding size and types of training and test data sets, unknown reproducibility/performance in real world settings (not to mention the cost and contraints of required fMRI/MEG machines), even though the newer LLMs and AI models can help ignore reconstruction errors. But with more money being thrown at these problems, and emerging interest for BCI in large profitable corporations, and the investigation of newer types of neuroscience-focused AI foundation models, the future will be interesting. Regarding the privacy issue, these systems record what the person is consciously imagining, at least in "normal" subjects, so perhaps less of a concern at least so far is my thinking, unless you accept the "memory download" potential some have been advocating. Without my rambling on too much, the most valuable initial use cases appear either for accessibility for physically challenged people, or new use cases we are yet to think. For example, for most people it is easier, cheaper, and far more reliable to use your hand to control a switch rather than to use brain signals for the foreseeable future, the advantage, should reliability improve is for applications requiring ultra-low latency (e.g. professional gaming), but reliability cannot be compromised. Sometimes truly good use cases for some new technologies can take decades to develop in the absence of substantial marketing budgets :)
I agree with all that you write. But it is difficult to gauge the speed of change in this digital era. I am still amazed that AI-generated photorealistic imagery has become commonplace just three decades or so after the emergence of Photoshop. Now that seems relatively slow.
True. And with the amount of (unprecedented) financial investment going into this space now, acceleration of progress seems to have good odds.
Dear Fred, Thank you for this piece, a lot to chew on. I have wanted a camera to be implanted in my forehead long ago when I realized I saw the photograph through my eyes not a viewfinder, and was lucky if the image stood still long enough for me to record it.