Locked In Syndrome - AI Breakthrough

[Rowls]

There's a lot of pseudo-scientific mumbo jumbo written about AI and the potential dangers it poses.

They might be, they might be overactive imaginations.

But forget all the noise and take a look at the link below - it's amazing:

https://twitter.com/rowancheung/status/ ... 5115919536

A lady who suffered a stroke and consequently locked-in syndrome is now able to express thoughts and "talk" via an avatar which is controlled through an implant in her brain. An AI interface decodes the lady's brain impulses and translates it into speech.

It's breathtaking. Well done to the research team. Little stories like this give us a tiny glimpse of how brilliant human beings can be. Hopefully it shine an aperture of positivity into your day and hopefully the technology can help those who suffer from this truly debilitating condition.

[FCBurnley]

It’s like most things. Amazing in the right hands and beyond dangerous in the wrong hands ( of which there are far too many )

[bobinho]

Don’t want this to come across badly in any way, but if she’s locked in, and this AI is translating her brain impulses into speech, how do we really know it’s getting it right?

[Rowls]

Don’t want this to come across badly in any way, but if she’s locked in, and this AI is translating her brain impulses into speech, how do we really know it’s getting it right?

I think it's still possible to communicate via blinking and to read her emotions.

I'm sceptical to know what level of intricacy this system is creating sentences for her but it appears to be a welcomed advance nontheless.

[DCWat]

We are only at the very beginning of what AI will deliver.

Potentially amazing and positive, in the hands of those that are searching for solutions across all areas of the worlds problems. Equally, worrying and potentially very dangerous in the wrong hands.

The bigger concern for me is how this technology can be kept in the right hands, I’m not sure it can, no matter how stringent governments are. Perhaps it won’t matter if the machines take over anyway.

[dougcollins]

Re read Asimov's robot trilogy recently.

More relevant than ever.

[Bosscat]

Skynet

[bobinho]

Careful boss... you'll end up being branded a lunatic conspiracy theorist if you start thinking anything different from all the positive AI spin the mainstream media is pumping out.

Lots of folk worried about it ending up in the wrong hands - people forgetting the 'wrong hands' will actually be the AI itself....

[morpheus2]

What a shame Steven Hawking is RIP now. Wonder if it could catch dreams too?

[Clive 1960]

That is great break through because the locked in syndrome after the stroke isn't pleasant for the patient and someone close to me had one it took a long time to get some recovery for her . Gary Parkinson had the same one as her , only a small percentage recover so this is a welcome boost.

[dibraidio]

Careful boss... you'll end up being branded a lunatic conspiracy theorist if you start thinking anything different from all the positive AI spin the mainstream media is pumping out.

Lots of folk worried about it ending up in the wrong hands - people forgetting the 'wrong hands' will actually be the AI itself....

I have to disagree. There are a lot of people in the industry who are extremely cautious about the big ethical questions of AI. There has been a lot of talk about chatgpt which is a large language model. A while back people were questioning why google were coming on this site with their bots, well, they were looking at the way that people exchange in "real" online conversation. Trying to understand natural language. Slightly worrying because who wants a chatbot who's going to start an arguement after just a handful of phrases?

One of the hot topics is goverance in LLMs (Large Language Models), ChatGPT has data from before 2021 in it's model because they've had to sort the data. You can't just let it loose on the internet, imagine if the bot comes back telling you that the last American election was rigged and that the earth is flat and created 6000 years ago.

A lot of the software in the realm of machine learning is open source but the LLM data is proprietary. A few years back IBM did some work on facial recognition using photos from Flickr. When people got upset about it they abandoned the project entirely because they want to be seen as "Good Tech" and not "Bad Tech". They want to respect people's rights when it comes to data about themselves. They are also building LLMs which don't contain foul language or any kind of bias. The models used are really important to allow chatbots to understand natural language and to communicate in a natural way with users.

The current targets for this kind of technology are the kind of jobs with repeatable tasks, first line telephone support for example. One thing that is interesting is that with an hour of recordings of someone's voice we can now simulate that person completely so for example you could have a chatbot that talks exactly like Winston Churchill with the same expressions, mannerisms and intonations. I recently heard an ai generated recording of Johnny Halliday singing La Boheme. A bit like having Freddy Mercury sing phantom of the opera, we can now generate incredibly convincing sound and images that never existed but are indistinguishable from reality. It would be incredibly easy to generate videos of Donald Trump being racist or sexist and cause a political storm using social media. This is the kind of realistic danger that we face. Fake news will become ever more real looking and we're not going to know what is true anymore.

Someone said to me when AI learns to read it'll be able to read every book that was ever written, that's already the case. We have commercially available technology to read a hand written letter, translate it to a text file and to understand the natural language and either act upon the request or send the text to the right person to deal with it. For a printed book it's even easier but then we come to the question of intellectual property.

It's a bit of a minefield to stay on the side of "good tech" but this is a real selling point in the AI market, companies don't want a potty mouthed bot answering the phone.