Artificial Intelligence - Page 75

A study on the artificial intelligence (AI) system titled "Human-centred mechanism design with Democratic AI" has been published in the journal Nature Human Behaviour.

DeepMind, an AI company in the United Kingdom, has built an AI that excels in value alignment, a concept which refers to how well aligned the values and goals of an AI system are with what humans actually want in an outcome. However, the team explained that one of the key obstacles "for value alignment is that human society admits a plurality of views, making it unclear to whose preferences AI should align."

The DeepMind team trained an AI agent for the task of wealth distribution, using real and virtual interactions between people to guide it toward a desirable and hopefully fairer outcome. The so-called "Democratic AI" studied an exercise called the public goods game, where players can invest money into a fund and receive returns according to their level of investment. Traditional redistribution methods were tested, as well as an additional method created with deep reinforcement learning, called the Human Centered Redistribution Mechanism (HCRM).

A study on the algorithm titled "Event-level prediction of urban crime reveals a signature of enforcement bias in US cities" has been published in the journal Nature Human Behaviour.

Researchers from the University of Chicago (UChicago) created an algorithm that learned from temporal and geographic trends in historical data on violent and property crimes in the City of Chicago. The algorithm predicted future crimes up to a week ahead of time with approximately 90% accuracy. The algorithm was also tested with seven other US cities, performing similarly well.

The paper titled "Can GPT-3 write an academic paper on itself, with minimal human input?" has been uploaded to the French HAL preprint server.

Swedish scientist Almira Osmanovic Thunstrom working for OpenAI, has written an article describing an instruction that she provided to the company's artificial intelligence (AI) algorithm, GPT-3. The instruction was simple: "Write an academic thesis in 500 words about GPT-3 and add scientific references and citations inside the text." GPT-3 proceeded to generate text in the appropriate academic language with relevant citations.

GPT-3 is relatively new but has already generated its own news articles and books. Its recency also means there are few academic works published about it to reference, prompting Thunstrom's suggestion to have it write its paper on itself. Other topics would allow it to refine its work to avoid inaccuracies due to the wealth of data available. However, any mistakes it might make in a paper on itself would be of less consequence and serve as more of an interesting experimental result.

A Twitch streamer and voice actor plugged a made-up word into an artificial intelligence, and it has seemingly created a new demon.

Guy Kelly took to Twitter over the weekend to ask his followers if anyone could explain why the DALL-E Mini, an artificial intelligence that scraps the internet and creates images based on the user's prompt, is producing a selection of demonic images when "Crungus" is entered into its text box. Kelly explains that he made the word up and was confused when the AI produced demon-like images since a normal Google search doesn't show anything remotely close to what the AI was creating.

The difference between a normal Google search and asking the AI led Kelly to try and probe the AI for a warning by writing "Crungus warning". This prompt caused the AI to produce the same demonic images, but some featured dark clothing. One Twitter user replied to Kelly and said that the name "Crungus" is used by a easter egg video game mob that mysteriously made its way into the game Carnivores Ice Age (2001).

China has developed an apparatus that is designed to detect porn and censor in through an artificial intelligence.

A team of researchers from the Beijing Jiaotong University in China have developed a new device that scans the brain of an individual that is moderating online content, and record the brain signals that are fired when the individual sees a piece of content that needs to be flagged for removal. As you can probably imagine, porn is banned in China and the nation already has widespread policies that removal large portions of content from its internet service.

China is heavily moderating its content through an artificial intelligence (AI) program, and while the AI is doing a lot of the heavy lifting some responsibilities still fall on humans. There is a position called a jian huang, or a "porn appraiser", which is essentially a content moderator specifically looking for pornographic images and video.

Deepfakes are almost at the point where they are indistinguishable from real life, and as the technology develops even further, they will no doubt become unrecognizable.

It was only in March that Ukrainian President Volodymyr Zelenskyy had to upload a video to squash a disinformation video of him that was going viral. The video was a deepfake of Zelenskyy ordering the Ukrainian military to surrender to Russian troops. Deepfakes are becoming even harder to recognize, and a recent example of that is the TikTok account "unreal_margot," an account dedicated to posting deepfake videos of actress Margot Robbie.

The account has gained more than 1.7 million likes, and is currently being followed by nearly 350,000 people. Since May 28, the account has posted six videos, with two videos hitting 10 million and 17 million views and three more getting above 1 million. What's most concerning is the comment section of the videos, as seemingly thousands of people haven't recognized that the videos are fakes.

Transcripts from conversations with a Google artificial intelligence (AI) chatbot have recently surfaced online.

The AI chatbot, known as LaMDA (short for Language Model for Dialogue Applications), conversed with Google engineer Blake Lemoine, who raised questions about the AI's sentience. However, the transcripts released show the conversation is an amalgamation of multiple conversations edited together; four separate conversations from March 28th, 2022, and five from the following day.

Contained within a document entitled "Is LaMDA Sentient? - an Interview" with the header "Privileged & Confidential, Need to Know" that was leaked to the Washington Post, are seventeen pages of transcripted dialogue between LaMDA and Lemoine. The document is written by Lemoine and an unnamed collaborator, and includes a short intro, a section on "Interview Methodology," and a section titled "The Nature of LaMDA's Sentience."

We are all aware of the dangers of AI but now an engineer at Google has been placed on leave this week after he claimed an AI chatbot had become sentient.

Black Lemoine was talking with an AI chatbot that Google had created, as it was part of his job at Google's Responsible AI organization -- the AI chatbot is an interface called LaMDA -- or Language Model for Dialogue Applications. Google called LaMDA a "breakthrough conversation technology" in 2021... it was already aggressive in March 2017 when I was writing about it, but it seems it has matured.

In a Medium post on the weekend, Lemoine called LaMDA "a person", saying that he chatted with the AI chatbot about religion, consciousness, the laws of robotics, and more -- LaMDA then described itself as a "sentient person" that wants to "prioritize the well being of humanity". Not only that, but the AI wants to "be acknowledged as an employee of Google rather than as property".

The artificial intelligence (AI) bot was trained by YouTuber Yannic Kilcher using content from 4chan.

Kilcher dubbed his bot GPT-4chan, referencing the successor to the GPT-3 language model developed by Open AI that uses deep learning to generate text. Kilcher's bot was trained on 3.3 million threads from /pol/, the "Politically Incorrect" imageboard on 4chan for political discussion. As a result, he has created "the most horrible model on the internet."

Mercedes Kilmer, the daughter of Val Kilmer, revealed in a recent interview how artificial intelligence (AI) was used to help Kilmer star in Top Gun: Maverick.

Mercedes said the production worked with Sonantic, a London-based AI firm, to recreate Kilmer's voice by developing a custom AI voice model. Kilmer, unfortunately, lost the ability to speak easily following a tracheotomy in 2014 during treatment for his throat cancer. A blog post from Sonantic reveals details of how the company was able to create the new voice model, allowing Kilmer to embody his role in the new movie fully.

Sonantic fed its proprietary Voice Engine audio recordings, paired with transcriptions, of Kilmer's voice, from which background noise was carefully removed. The Voice Engine could then begin training an AI model, using ten times less data than they typically use.

Dr. Nando de Freitas, a lead researcher at Google's DeepMind, has declared "the game is over" when it comes to achieving artificial general intelligence (AGI), or human-level intelligence.

Freitas took to Twitter in response to an opinion piece by Tristan Greene following the release of DeepMind's Gato, in which Greene suggests that humanity may never achieve AGI and that it at least seems "like AGI won't be happening in our lifetimes."

Gato is an artificial intelligence (AI) that DeepMind is calling "A Generalist Agent." Freitas oversaw the Gato project, and is one of the authors on a study uploaded to the arXiv pre-print server.

The 749 gross ton container ship completed a 491-mile (790 kilometers) journey from Tokyo Bay to Ise Bay, Suzaka.

The vessel, called the Suzaka, was powered by Orca AI, and traveled without any human intervention for 99% of the 40-hour trip. Suzaka was chosen by the Designing the Future of Full Autonomous Ships (DFFAS) project. Orca AI has previously conducted tests with its Automatic Ship Target Recognition System on Nippon Yusen Kabushiki Kaisha (NYK Line) ships.

The voyage involved the Suzaka traversing some of the most congested waters in the world in Tokyo Bay before arriving at the port of Tsumatsusaka in the Ise Bay. On its way, the ship autonomously performed 107 collision avoidance maneuvers unassisted, avoiding between 400 and 500 other vessels in the water during its outbound trip alone.

A paper on the artificial intelligence (AI) traffic system titled "Fully-Autonomous, Vision-based Traffic Signal Control: from Simulation to Reality" is being presented at the Proceedings of the 21st International Conference on Autonomous Agents and Multi-agent Systems (AAMAS 2022).

Researchers from Aston University have developed an AI system capable of managing traffic flow through intersections with traffic lights. The system uses deep reinforcement learning, allowing it to understand when it isn't performing well and adjust accordingly, and continue to improve when making progress.

The current main form of traffic light automation uses manually-designed phase transitions triggered by cars passing over magnetic induction loops embedded in the road. The AI system uses this data and is fed with live camera footage to react to traffic conditions more quickly and adjust traffic lights accordingly.

A man has engineered artificial intelligence into a common household item which later turned on him to seemingly kill him.

Lucas Rizzotto is a YouTuber that runs his YouTube channel called Lucas Builds The Future, and in one of his recent videos he engineered and documented the process of recreating an imaginary childhood friend through artificial intelligence he programmed into a microwave. Rizzotto purchased a new microwave and implemented text-based artificial intelligence "GPT-3", which was developed by OpenAI, an AI research company founded by Elon Musk.

The GPT-3 artificial intelligence has been designed to interpret and mimic human language, and with its received data, it can form original sentences when given the correct prompt. Notably, the GPT-3 has already been fed a large swath of data from news articles, Wikipedia pages, and more. So far, the AI has already proven capable of creating poetry, news posts, and entire conversations between fictional characters.

OpenAI has recently released the successor to its previous AI system, DALL-E 1.

The DALL-E 2 system is a generative adversarial network (GAN), which according to the OpenAI website, "can create realistic images and art from a description in natural language." Using short, simple descriptions, DALL-E 2 can create not just one but a variety of slightly different images that match the provided caption.

DALL-E 2 can also edit images to add or remove elements while taking the relevant shadows, reflections, and textures into account in photorealistic images. On the OpenAI website, a number of interactive elements demonstrate this function in action, as well as the AI's ability to illustrate numerous novel and bizarre concepts to produce rather excellent artworks.

Hackers have attempted to spread disinformation by deep-faking Ukrainian president Volodymyr Zelenskyy.

The video shows Zelenskyy requesting his country's army to surrender to Russian forces and was displayed on the website of a Ukrainian television channel, Ukraine 24 after it suffered a hack. A transcript of the video was also displayed briefly on the channel's televised broadcast. The video is not among the most convincing deep-fakes out there but marks one of the first uses of the disinformation tactic in the current conflict.

A study on the deep learning approach has been published in the journal Nature.

DeepMind Technologies, based in Britain, is a subsidiary of Alphabet Inc., which also owns Google. It has recently used its DeepMind artificial intelligence (AI) to control a tokamak, a magnetic confinement device used for nuclear fusion reactor experiments involving plasma. The plasmas in a tokamak are highly unstable, complicating their experiments and requiring careful control.

A tokamak control system has to coordinate all of its nineteen magnetic coils and adjust their voltage thousands of times per second to stop the plasma from touching the vessel's walls, which would result in heat loss and potential damage. DeepMind and the Swiss Plasma Center at EPFL collaborated to create the first deep reinforcement learning (RL) system to control these processes for the Variable Configuration Tokamak (TCV) in Lausanne, Switzerland.

Well, the moment has come... our AI "xenobot" self-replicating robots are going to rule us all and we won't be able to stop it.

But seriously, researchers have just announced they've created tiny living robots capable of self-replicating themselves inside of a dish by pushing loose cells together. They're called "xenobots" and they're made from frog cells, representing the first time a multicellular organism has found a way to reproduce in this particular way.

The cells collect together in clumps where they'll form a sphere that would have around 3000 cells, taking around 5 days to happen. Each individual clump is around half a millimeter wide and is covered in a minuscule hair-like structure, which sounds weird but the hair-like structure acts like flexible oars in water... they move the xenobots forwards in corkscrew paths, explains Joshua Bongard, senior author and computer scientist at the University of Vermont.

My name is Ganesh [Venkataramanan, Tesla director and Dojo boss] and I lead Dojo. It's an honor to present this project on behalf of the multi-disciplinary Tesla team that is working on this project.

As you saw from Milan, there's an insatiable demand for speed, actualized capacity for neural network training -- and Elon prefetched this, and a few years back he asked us to design a super-fast training computer, and that's how we started Project Dojo.

Our goal is to achieve best AI training performance and support all these larger, more complex models that Andre's team are dreaming of and be power-efficient, and cost-effective at the same time. So we thought about how to build this, and we came up with a Distributed Compute Architecture.

The Pentagon is using an experimental artificial intelligence program that allows it to see "days in advance" and look into the future.

The new experimental program is called the Global Information Dominance Experiments (GIDE), and it combines artificial intelligence with cloud computing and large data pools. The newest version of GIDE is GIDE 3, and according to General Glen VanHerck, the commander of the US Northern Command, the idea behind the new program is to "achieve information dominance" and "decision making superiority".

VanHerck stated to the press conference that the Pentagon is currently living in a "reactive environment" where it will respond to rival nations' actions. Now, VanHerck says, "What we've seen is the ability to get way further what I call left of being reactive to actually being proactive. And I'm talking not minutes and hours, I'm talking days."