Artificial Intelligence & Robotics

[nudge+]

9 hours ago, Mel81x said:

For @Eco and @nudge

https://arstechnica.com/science/2020/11/when-ai-sees-a-man-it-thinks-official-a-woman-smile/

Fascinating read on how the AI used biased-data sets to build model cards for pictures it was presented. 

Another example of biased training datasets resulting in biased algorithms... It's been such a commonly reported problem, I wonder if anyone is doing anything at all to actually improve it, or do they just scrap it once they realise how flawed it is, and then repeat the same mistakes again the next time...

[Mel81x+]

25 minutes ago, nudge said:

Another example of biased training datasets resulting in biased algorithms... It's been such a commonly reported problem, I wonder if anyone is doing anything at all to actually improve it, or do they just scrap it once they realise how flawed it is, and then repeat the same mistakes again the next time...

I think this is where the learning part of the AI really needs to take the next step. Its all well and good to say you have a system which reads data and creates methodologies to create pathways for interpretation but its a whole other story to listen to an externalized source (machine/human) and then use that input to auto-correct behavior. Obviously these machines havent been around for decades or they may have learned but as you rightly say who gives these machines a chance to actually learn from their mistakes? Most of them are fancy data accumulators whose only purpose is to generate actionable ROI on whatever company is building them which in essence isn't true AI.

[CaaC (John)+]

Not sure if this is the right forum for this?

 

Quote

Next-gen drones could bee like this

Agility and perception the keys to navigating tight spots

Researcher Sridhar Ravi with a bumblebee-inspired drone. Credit: Lannon Harley / UNSW Canberra

An Australian-led team says it has uncovered the secret to the agile flight of the bumblebee, which could influence the next generation of drone technology.

To understand how the bees navigate their environments, researchers created an obstacle course consisting of a tunnel separating their hive from their foraging area, blocked by a series of gates with different-sized holes.

A high-speed camera followed them whizzing through without a hitch, demonstrating a keen awareness of the layout of their surroundings and an accurate sense of their own body size and capacity for movement.

This is the first time that such evidence has been seen in flying invertebrates.

“Previous research had indicated that complex processes, such as the perception of self-size, were cognitively driven and present only in animals with large brains,” explains lead author Sridhar Ravi, from UNSW Canberra.

“However, our research indicates that small insects, with an even smaller brain, can comprehend their body size and use that information while flying in a complex environment.”

The project was a collaboration between UNSW Canberra, Bielefeld University and the Max Planck Institute in Germany, and Brown University and the University of California in the US. The findings are published in the journal PNAS.

Bumblebees were chosen as a model because of their body size varies greatly, even within a colony, and they have a proven ability to move through densely-cluttered environments to forage.

The researchers observed bees of all sizes using a process called “lateral peering”, in which they paused in front of an aperture and flew side-to-side to give themselves time to assess it, before choosing how to adjust their posture to fly through. The narrower the gap, the longer the bees took to scan it, with the largest bees taking the longest time.

In all 400 flights observed, the bees made it successfully through every gap, though their wings occasionally clipped the edges of smaller gaps.

In extreme cases, bees were even observed turning 90 degrees to fit through a gap smaller than their wingspan, like humans turning their shoulders sideways to fit through a narrow doorway.

“We were amazed to see that in some instances, the bumblebees reoriented themselves sideways to fly through gaps they were unable to attempt head-on,” says Ravi. “The dexterity of these insects has really got us thinking about what other secret bee behaviours we could unlock.”

This remarkable spatial- and self-awareness may be widespread in insects, and it may find applications in future technologies such as drones and autonomous vehicles, helping them to meet the challenges of navigating real-world environments.

“Insects are fantastic models for robots because they have exceedingly small brains and yet they’re able to perform overly complex tasks,” Ravi says.

“Over thousands of years, nature has coded insects with some amazing attributes. Our challenge now is to see how we can take this and apply a similar coding to future robotic systems, enhancing their performance in the natural world.”

https://cosmosmagazine.com/news/next-gen-drones-could-bee-like-this/

 

[Mel81x+]

3 hours ago, CaaC (John) said:

Not sure if this is the right forum for this?

Fascinating article. Now, if we could that self-awarness of body size from flying invertebrates and get people to stop trying to fit into clothes that are too small for them I'll call it a success. 

[CaaC (John)+]

Quote

AI 'doctor's assistant' among projects given £20m

A computerised system that advises doctors on the best treatment to give a patient is among the artificial intelligence projects that have been awarded £20m by the British government.

The money will be shared between 15 AI projects being run at UK universities.

It's been done as part of the Turing fellowship scheme, named after maths genius and AI pioneer Alan Turing.

The projects help the UK meet some of today's most pressing challenges.

These include developing more effective ways of treating cancer and supporting efforts to tackle climate change.

AI describes software systems that make decisions usually requiring human expertise.

• How safe can artificial intelligence be?
• Why hasn’t AI changed the world yet?
• How artificial intelligence may be making you buy things
• How can AI help to fight against coronavirus?

It has the potential to transform the way we live, allowing tasks to be performed faster and more accurately than they are by people.

The doctor's assistant, or clinical colleague, is a project being led by Professor Aldo Faisal, of Imperial College London. It would be able to recommend medical interventions such as prescribing drugs or changing doses in a way that is understandable to decision-makers, such as doctors.

This could help them make the best final decision on a course of action for a patient. This technology will use "reinforcement learning", a form of machine learning that trains AI to make decisions.

The aim is to relieve the pressures and workload on doctors and clinicians. But similar systems could be used in sectors such as aerospace or energy, where there is a need for decision-making support.

Another of the projects will aim to use AI to spot cancer before it can develop and spread in the body. It's being led by Prof Christopher Yau at the University of Manchester.

Prof Yau said it would involve using the vast amounts of information produced by genomic sequencing - the ability to sequence the full complement of DNA contained in the nuclei of human cells, including cancer cells.

If successful, it will enable medical experts to track cancer more accurately and help them decide at an earlier stage what treatments patients require. This would increase the chances of saving lives as treatment is usually more successful when given earlier.

Another project, led by Dr Antonio Hurtado, from the University of Strathclyde, aims to meet the growing demand across the UK economy to process large volumes of data fast and efficiently - while minimising the energy required to do so.

His AI system will use laser light, similar to technology used in supermarket checkouts, to perform complex tasks at ultra-fast speeds. It could be used in areas such as weather forecasting and processing images for medical diagnostics.

Being able to perform these tasks at lightning speed, with minimal energy consumption, could help to transform industries such as energy, healthcare and finance, improving efficiency, while helping the UK to meet its net-zero ambitions by 2050.

Science Minister Amanda Solloway said: "The UK is the birthplace of artificial intelligence and we have a duty to equip the next generation of Alan Turings with the tools that will keep the UK at the forefront of this remarkable technological innovation.

"The inspirational fellows we are backing today will use AI to tackle some of our greatest challenges head-ons, transforming how people live, work and communicate, cementing the UK's status as a world leader in AI and data."

https://www.bbc.co.uk/news/science-environment-55099620

 

[nudge+]

Thought this was worth sharing... AI has made a significant breakthrough in the protein folding problem that has kept scientists busy for almost 50 years. The implications of this for life and medical science, especially drug discovery and development, are huge.

https://www.nature.com/articles/d41586-020-03348-4

And a video to go along with it:

 

 

[CaaC (John)+]

Quote

DeepMind's AI agent MuZero could turbocharge YouTube

DeepMind's latest AI program can attain "superhuman performance" in tasks without needing to be given the rules.

Like the research hub's earlier artificial intelligence agents, MuZero achieved mastery in dozens of old Atari video games, chess, and the Asian board games of Go and Shogi.

But unlike its predecessors, it had to work out their rules for itself.

It is already being put to practical use to find a new way to encode videos, which could slash YouTube's costs.

"The real world is messy and complicated, and no-one gives us a rulebook for how it works," DeepMind's principal research scientist David Silver told the BBC.

"Yet humans are able to formulate plans and strategies about what to do next.

"For the first time, we actually have a system which is able to build its own understanding of how the world works and use that understanding to do this kind of sophisticated look-ahead planning that you've previously seen for games like chess.

"[It] can start from nothing, and just through trial and error both discover the rules of the world and use those rules to achieve kind of superhuman performance."

FULL REPORT

 

[Mel81x+]

https://arstechnica.com/science/2020/12/google-develops-an-ai-that-can-learn-both-chess-and-pac-man/

Now, call me cynical but whenever I see AI try to play Chess or something as 'simple' as Pac-Man a part of me wonders how much AI is involved versus classic problem-solving. Lets think about it this way and use both games.

1. Chess - Finite position/vector-based game. Only so many squares, only so many moves. Only so many permutations allowing a machine to see to what we may consider infinity till the next move is played. Realistically the only time the AI might kick in is if it learns the other player and then tries to adapt its game but even then the only real move that can trip it properly is the first move. So is it really AI? Not really but lets call it AI because that's cool and in-line with how tech is hyped. IBMs big show and its failure when a human beat it wasn't AI it was just poor problem solving and non-adaptability because lets face it if the AI was really present no human would have been able to touch it, period. 
2. Pac-Man - You'd imagine a game with the limitations of space like Pac-Man would be just like Chess then you realize that the ghosts have finite moves they can execute to achieve goals and more importantly they are non-adaptive so really this is once again problem-solving but with a slight twist. Your problem solving using four moving points to determine the best approach. So is it really AI once again? No. When does it become AI? When the ghosts stop becoming predictable and also use AI to beat the AI that is running Pac-Man. Then lets add that RAND (random number generator) so famously used in the 8-bit era to create the illusion that the patterns you were seeing were truly random. They really aren't, so once the AI pattern matches the seed generator it can go forever without flinching because the only thing the game has to ramp difficulty at that point isn't the ghosts its the speed of the game and with another machine playing on the other side that is never going to be a problem ever. Might be fun to watch though. 

In the end I think we generously throw this word AI around at things that aren't true AI. Adaptation is a key-factor yes but so is the process of elimination to determine a play-style that best mimics something that has the smarts to play the game. I also think games are the worst way to describe AI because they are finite in the way they can be played with some exceptions of course. Some of the best forms of AI are in language understanding and cultural adaptation with assimilation. 

[nudge+]

11 minutes ago, Mel81x said:

https://arstechnica.com/science/2020/12/google-develops-an-ai-that-can-learn-both-chess-and-pac-man/

Now, call me cynical but whenever I see AI try to play Chess or something as 'simple' as Pac-Man a part of me wonders how much AI is involved versus classic problem-solving. Lets think about it this way and use both games.

1. Chess - Finite position/vector-based game. Only so many squares, only so many moves. Only so many permutations allowing a machine to see to what we may consider infinity till the next move is played. Realistically the only time the AI might kick in is if it learns the other player and then tries to adapt its game but even then the only real move that can trip it properly is the first move. So is it really AI? Not really but lets call it AI because that's cool and in-line with how tech is hyped. IBMs big show and its failure when a human beat it wasn't AI it was just poor problem solving and non-adaptability because lets face it if the AI was really present no human would have been able to touch it, period. 
2. Pac-Man - You'd imagine a game with the limitations of space like Pac-Man would be just like Chess then you realize that the ghosts have finite moves they can execute to achieve goals and more importantly they are non-adaptive so really this is once again problem-solving but with a slight twist. Your problem solving using four moving points to determine the best approach. So is it really AI once again? No. When does it become AI? When the ghosts stop becoming predictable and also use AI to beat the AI that is running Pac-Man. Then lets add that RAND (random number generator) so famously used in the 8-bit era to create the illusion that the patterns you were seeing were truly random. They really aren't, so once the AI pattern matches the seed generator it can go forever without flinching because the only thing the game has to ramp difficulty at that point isn't the ghosts its the speed of the game and with another machine playing on the other side that is never going to be a problem ever. Might be fun to watch though. 

In the end I think we generously throw this word AI around at things that aren't true AI. Adaptation is a key-factor yes but so is the process of elimination to determine a play-style that best mimics something that has the smarts to play the game. I also think games are the worst way to describe AI because they are finite in the way they can be played with some exceptions of course. Some of the best forms of AI are in language understanding and cultural adaptation with assimilation. 

I think the fact that MuZero learns without being given the rules by developing its own internal dynamic model of the environment and optimising within that model, and is also non-specific enough to teach itself new games is pretty neat though... Kind of similar to how we (humans) learn things without reading the rulebook first. Now, I agree with you that it's still more of a statistical model which mimics the most effective play, but while it might not be true AI, the progress reinforced learning has made in such a short period of time is still very impressive. Hopefully this will lead to it mastering more complex multi-agent games by demonstrating its understanding of causal connections between a multitude of different actions instead of relying on brute forcing solutions.

[Mel81x+]

7 hours ago, nudge said:

I think the fact that MuZero learns without being given the rules by developing its own internal dynamic model of the environment and optimising within that model, and is also non-specific enough to teach itself new games is pretty neat though... Kind of similar to how we (humans) learn things without reading the rulebook first. Now, I agree with you that it's still more of a statistical model which mimics the most effective play, but while it might not be true AI, the progress reinforced learning has made in such a short period of time is still very impressive. Hopefully this will lead to it mastering more complex multi-agent games by demonstrating its understanding of causal connections between a multitude of different actions instead of relying on brute forcing solutions.

That is the most fascinating part and I suppose I am being harsh on the science of AI which is at its heart trying to mimic something organic the best way it knows how, by building models it can learn off. I too am curious to see how well it adapts to games it knows very little about in the future but i suspect we're a while away from that right now.

[CaaC (John)+]

Quote

Tightening the dragnet on Denisovans

AI tech tries to sniff out hidden signatures of ancient humans in modern DNA.

Researchers have identified some prime suspects in the biggest ‘who done it’ mystery in human evolution: who were the Denisovans?

In a study published in Nature Ecology and Evolution, a team – led by population geneticist João Teixeira at the University of Adelaide in Australia – has attempted to nail down the identity of these enigmatic ancient humans by using AI to probe deep into the DNA of modern people of southeast Asia.

“Denisovans are making people rethink what they thought they knew,” says Teixeira, who collaborated with Murray Cox at Massey University, New Zealand; Guy Jacobs at the University of Cambridge, UK; Chris Stringer at London’s Natural History Museum; and Kris Helgen at the Australian Museum in Sydney.

Denisovans are known only from a few sparse remains, including DNA from 50,000-year-old Siberian finger bone and teeth, as well as collagen proteins from a 160,000-year-old jaw fragment in Tibet. Intriguingly, these bits of bone and teeth don’t match any of the known fossils in the human family tree.

In 2010, DNA extracted from the finger bone confirmed that this is a completely new species (or subspecies – taxonomists can’t agree). But Denisovans aren’t just some curious relic of our past – we still carry significant chunks of their DNA today, which suggests they interbred with modern humans as recently as 55,000–30,000 years ago. Genetic studies reveal very little Denisovan DNA in modern Europeans and Asians (less than 0.1%), but high percentages (around 4%) in New Guinea and Australia and the Mamanwa from the Philippines, people with ancestry from the traditional hunter-gatherers of the Asia-Pacific. (For comparison, Neanderthal DNA is found in all populations outside Africa at 1–3%.) This suggests the most recent trysts between Denisovans and modern humans took place in New Guinea and Australia.

So who exactly were these trans-Eurasian trekkers? And why haven’t we found their remains in south-east Asia? Or is it possible we’ve misidentified existing fossil humans – and some might actually be the mystery “southern” Denisovans? The problem is, none of the key fossil suspects are forthcoming with their DNA: the tropics are unkind when it comes to preservation.

As a workaround, this new study deployed AI to sniff out cryptic signatures of ancient humans in the DNA of modern people of island southeast Asia (ISEA).

Until recently, the line-up of fossil Denisovan suspects in Southeast Asia would have been limited to the hefty Homo erectus of Java, which had a brain size approaching that of modern humans, left Africa around 1.9 million years ago and roamed Java from 1.5 M years to 108,000 years ago. In 2004 the dwarf species H. floresiensis joined the line-up of suspects. Known to have lived on the island of Flores as recently as 60,000 years ago, individuals were a metre tall and had a brain capacity of 426 cubic centimetres, about one third that of a modern human. In 2019, the equally dwarfish H. luzonensis was added to the list; fossil remains from the island of Luzon in the Philippines reveal that this hominid stood around a metre tall and existed over a similar time span.

These three species are termed “super-archaics”. When it comes to their spot in the hominin family tree, most anthropologists place them on a branch that split from our line two million years ago.

To test if any of these super-archaics might be Denisovans, Teixeira and the team trained an AI to use a Hidden Markov Model to “walk” along with the DNA code, sniffing for two-million-year-old DNA. Thanks to the efforts of Herawati Sudoyo at the Eijkman Institute for Molecular Biology in Jakarta, who painstakingly gathered tissue samples from isolated populations ranging from tiny islands to the remote highlands of New Guinea, the AI was able to probe the genomes of 200 people from ISEA – populations that seem to have acquired their Denisovan DNA as recently as 30,000 years ago. This needle-in-a-haystack search method can detect traces of super-archaic code that represent 0.1% of the DNA – “one in a thousand ancestors”, emphasises Teixeira.

The first step was to mask off Neanderthal and Denisovan signatures, as well as any variant signatures also found in African populations. That sensitised the algorithm to see any new signatures that had arisen in ISEA.

The verdict?

Inconclusive.

A faint whiff of two-million-year-old super-archaic DNA was identified, but it wasn’t strong enough to convince the authors this was introduced by a hominin down under. It may have been a “methodological artefact”: a leftover signature from the mingling between Denisovans and a super-archaic in the northern hemisphere, possibly H. erectus – a finding which others like Cornell University’s Melissa Hubisz have reported.

Either way, the authors agree there is no conclusive evidence for a new super-archaic signature in people of ISEA.

University of Wisconsin anthropologist John Hawks, who was not involved in the study, finds it a convincing piece of work – especially since previous reports did suggest such signatures in Indian and Asian populations.

“The search for super-archaics is a rich target,” he says. “It took someone using more modern methods that are not easily tricked.”

But if this study was inconclusive, where does this leave the hunt for the Denisovans?

The authors are somewhat divided. Most say the evidence does not support the possibility that the island pygmies or hulking H. erectus are Denisovans. One suggestion is to keep searching the little-explored caves of ISEA for the remains of Denisovans. Sulawesi is the hot favourite. It has stone tools dating back 200,000–100,000 years ago, as well as the world’s oldest cave paintings. 

A convincing fossil candidate should look more archaic than modern humans but not as archaic as the island hobbits – and it should have hung around till the moderns arrived about 50,000 years ago.

But others have not entirely let the line-up’s shady characters off the hook.

The ruling assumption was that the island hominins and H. erectus must all be super-archaic – in other words, their feature set is so ancient that they must have been travelling a separate track to modern humans for the past two million years. But that assumption could be flawed.

Perhaps the weird island hominids are not as super-archaic as they seem.

“Evolution goes crazy on islands,” says co-author Kris Helgen. Small founding populations and extreme conditions fire up the evolutionary engine – perhaps a mere 100,000 years ago Denisovans made their way to the island and not only shrank but also produced throwbacks to a more ancestral state.

It’s possible that H. erectus evolved in unanticipated ways, too. Traditionally viewed as travelling a separate track to modern humans for over two million years, not everyone buys that theory. Previous research suggested that H. erectus in Java and China was modernising over its two-million-year stint in Asia through interbreeding with newer hominins roaming Eurasia.  

It’s possible that a modified form of H. erectus – like the 108,000-year-old population found buried in the banks of the Solo river near Ngandong, Java – might be Denisovans.

“We might have to rethink H. erectus,” said Teixeira.

Hawks agrees: “My hypothesis is that its Ngandong.”

https://cosmosmagazine.com/history/palaeontology/tightening-the-dragnet-on-denisovans/

 

[nudge+]

@Mel81x have you read that pre-print paper by a group of theoretical physicists working for Microsoft, which describes our universe as an algorithm that’s continuously learning about itself? 

[Mel81x+]

15 hours ago, nudge said:

@Mel81x have you read that pre-print paper by a group of theoretical physicists working for Microsoft, which describes our universe as an algorithm that’s continuously learning about itself? 

Its odd you say this because I watched a science video about population growth recently lol and its this.

Do you have a link for the algorithm you're talking about? I can see references to the first Friedmann equation but nothing else.

Edited April 12, 2021 by Mel81x

[nudge+]

8 hours ago, Mel81x said:

Its odd you say this because I watched a science video about population growth recently lol and its this.

Do you have a link for the algorithm you're talking about? I can see references to the first Friedmann equation but nothing else.

Wow, that population growth graph is crazy... It's weird how from generation to generation the population seems to fluctuate between two extremes when the average number of offspring is above three, and how the population growth of any generation is basically randoms. Fractals make me dizzy, haha.

This is the paper I was talking about> https://arxiv.org/abs/2104.03902 
It's 80 pages in total, I haven't read it myself yet...

[Mel81x+]

1 hour ago, nudge said:

Wow, that population growth graph is crazy... It's weird how from generation to generation the population seems to fluctuate between two extremes when the average number of offspring is above three, and how the population growth of any generation is basically randoms. Fractals make me dizzy, haha.

This is the paper I was talking about> https://arxiv.org/abs/2104.03902 
It's 80 pages in total, I haven't read it myself yet...

I have my weekly reading now haha. And yes Fractals make me go a bit crazy too when they are put into equations.

[nudge+]

50 minutes ago, Mel81x said:

I have my weekly reading now haha. And yes Fractals make me go a bit crazy too when they are put into equations.

I'll try to read it this week, too... Let me know what you think afterwards!

[Mel81x+]

1 minute ago, nudge said:

I'll try to read it this week, too... Let me know what you think afterwards!

Most certainly.

[Azeem]

In case they ever build an evil clone of me there is a secret about me which only me and very selective people know, they would have no chance cloning that

[Mel81x+]

3 hours ago, McAzeem said:

In case they ever build an evil clone of me there is a secret about me which only me and very selective people know, they would have no chance cloning that

If Neuralink ever becomes a reality theres a very slim chance they wont know.

[CaaC (John)+]

Quote

Is AI sexist and racist?

From facial recognition to digital assistants - AI is all around us

We all use facial recognition to unlock our phones. And we all view online content automatically suggested to us. But some of us have rather more success with artificial intelligence (AI) than others.

A study of face recognition AIs discovered that systems from leading companies IBM, Microsoft and Amazon misclassified the faces of Oprah Winfrey, Michelle Obama and Serena Williams while having no trouble at all with white males.

Even the voices of digital assistants such as Cortana or Google Assistant have female voices by default, perhaps unconsciously reinforcing the stereotype of female subservience in the minds of millions of users.

The bias of these AIs is caused by the fact that the current designers of most AIs are largely white males in their 20s and 30s without disabilities. They’re generally people who grew up in high socioeconomic areas, often with similar educational backgrounds.

Perhaps unsurprisingly, the resulting AIs are created and educated using narrow and biased datasets that are unrepresentative. For instance, a US government dataset of faces collected for training AIs contained 75 per cent men and 80 per cent lighter-skinned individuals. There’s nothing deliberate about this – the AI developers simply didn’t notice because they had no experience of diversity themselves.

Thankfully the tide is turning, and today most major tech companies are trying to identify unwanted biases and eradicate them from our technologies.

https://www.sciencefocus.com/science/is-ai-sexist-and-racist/

 

[Mel81x+]

https://arstechnica.com/gaming/2021/05/robert-de-niro-speaks-fluent-german-in-taxi-driver-thanks-to-ai/

@nudge how did the AI do according to you?

[CaaC (John)+]

Quote

Podcast: How AI and androids could shape the music of the future

Prof Nick Bryan-Kinns discusses whether artificial intelligence is the new frontier for the music industry.

While Daft Punk may have sadly split, machine-created music may be about to skyrocket in popularity. Not only are artificial intelligence neural networks now capable of creating original melodies, but scientists are also developing robots capable of playing – and improvising – live music.

So, will AI and androids soon top the charts? And could they even replace human musicians entirely?

On this week’s episode of the Science Focus Podcast, Prof Nick Bryan-Kinns, director of the Media and Arts Technology Centre at the Queen Mary University of London, joins staff writer Thomas Ling to explain groundbreaking new music technology.

https://www.sciencefocus.com/future-technology/podcast-how-ai-and-androids-could-shape-the-music-of-the-future/

 

[nudge+]

15 hours ago, Mel81x said:

https://arstechnica.com/gaming/2021/05/robert-de-niro-speaks-fluent-german-in-taxi-driver-thanks-to-ai/

@nudge how did the AI do according to you?

Sorry, I read this early in the morning, then fell asleep again and forgot all about it

Well... It's clearly not flawless  Robert de Niro's face when he's "talking German" just looks plain weird, that jaw especially.. same with Nicholson in French, thought that also looked weird. Tom Cruise, on the other hand, looked very passable. I have no doubts they will improve the technology in the future, although I hope they don't as a) I prefer watching original facial expressions and hearing original language, and b) this is soooo going to be misused for all the wrong reasons. I fucking hate deepfakes, think we're stepping into a very dangerous zone here. 

[Mel81x+]

9 hours ago, nudge said:

Sorry, I read this early in the morning, then fell asleep again and forgot all about it

Well... It's clearly not flawless  Robert de Niro's face when he's "talking German" just looks plain weird, that jaw especially.. same with Nicholson in French, thought that also looked weird. Tom Cruise, on the other hand, looked very passable. I have no doubts they will improve the technology in the future, although I hope they don't as a) I prefer watching original facial expressions and hearing original language, and b) this is soooo going to be misused for all the wrong reasons. I fucking hate deepfakes, think we're stepping into a very dangerous zone here. 

I thought about deep fakes and then i said to myself "its already happening and does that mean we stop advancing?" Then I thought it would be cool to have someone like DeNiro tell me the news so why not? haha. 

[CaaC (John)+]

AI breakthrough could spark medical breakthrough.

Artificial intelligence has been used to predict the structure of almost every protein made by the human body.

https://www.bbc.co.uk/news/science-environment-57929095