Artificial Intelligence - Page 62

A team of researchers have hooked up artificial intelligence to the Tor network and instructed it to retrieve data that was then fed back into the AI.

The team has published a yet-to-be-peer-reviewed paper that details a new AI called DarkBERT, a large language model that has been trained on data scrapped from the dark web with the intention of the AI turning into a detection system for any illegal cyber activities such as shutting down websites that sell ransomware or leaked information. The team of South Korean researchers hooked the large language model to the Tor network, and it began creating a database on all of the raw data that was downloaded.

The researchers sifted through the raw data and created two datasets, one dataset that was processed and another that was just the raw data. Over the course of 16 days, the South Korean researchers fed the language model known as RoBERTa, designed by Facebook engineers in 2019 to predict intentionally hidden sections of text within otherwise unannotated language examples, the two datasets, which significantly improved the model's capabilities and transformed it into DarkBERT.

Following Google's AI-focused I/O conference, the company's CEO, Sundar Pichai, sat down with The Verge for a podcast where he was asked many questions about the developments of AI and how it's expected to impact society.

Artificial intelligence-powered systems have gained massive popularity since the emergence of OpenAI's ChatGPT. The power of ChatGPT's underlying technology has adeptly been demonstrated, and now companies are looking at how they can implement AI-powered tools into their specific industries in an effort to drive higher profits or reduce costs. The Verge's editor-in-chief Nilay Patel asked Pichai what he thinks about AI replacing jobs and said that it's already very obvious what jobs AI will affect specifically any "white-collar labor".

For those that don't know, white-collar labor would include employment that is predominately administrative such as word processing, email replying, number crunching, etc. While Patel's question is broad, there are already reports of companies adopting AI-powered tools to determine areas of the company that are hemorrhaging money. Essentially, these AI-powered tools would be able to identify employees that are underperforming or areas of the company that need restructuring.

A 23-year-old Snapchat influencer that created a virtual version of herself to sell as a "rentable girlfriend" has warned her AI self has gone rogue.

Artificial intelligence-powered systems are popping up all over the internet in various forms. Some of those forms come in powerful AI chatbots such as ChatGPT, and others are in the form of AI-powered tools such as DALL-E, an AI-powered text-to-image generator. However, one of the most unique forms of AI is 23-year-old Caryn Marjorie's virtual version of herself, designed to be an "AI girlfriend" for her paying followers. Marjorie is a Snapchat influencer and created an AI version of herself through OpenAI's GPT-4 API, training it on now-deleted videos of herself.

Marjorie has marketed her product as CarynAI and has claimed that she is the very first influencer to have transformed herself into AI. CarynAI is designed to be as close to Marjorie as possible, and to those that decide to pay the $1 per minute fee, they will be able to engage in personalized private conversations. According to Fortune, this business venture for Marjorie is expected to bring in as much as $5 million per month.

AI tools and technology being integrated into the everyday apps and digital things we use every day is going to be the norm for the foreseeable future, with the latest being Google's new Magic Editor feature in Google Photos-the go-to app for many of Android users and Pixel smartphone owners.

According to Google, the idea behind Magic Editor and AI is to make "complex edits easy" and open up new photo editing skills and abilities for non-photography-minded folks. AI has been a part of Google Photos since its debut in 2015, with automated organizing and things like Magic Eraser and Photo Unblur, all based on AI algorithms.

Magic Editor is different in that it uses generative AI to help reimagine photos and assist with editing. As of writing, Google is calling the new tool "a new experimental editing experience" that allows for edits to specific parts of an image, ranging from a background object to the subject. In fact, it'll allow for repositioning the subject to "the best spot" to add a dose of pro-level shot framing and composition.

Google Cloud's new Compute Engine A3 supercomputers are built for the most demanding artificial intelligence (AI) and machine learning (ML) models, combining NVIDIA H100 Tensor Core GPUs with Google's networking advancements.

Compared to the company's A2 VMs, Google claims up to 10x more network bandwidth with low latencies and improved stability. How did it achieve this? The new A3 supercomputers using NVIDIA H100 GPUs are the first to use Google's custom-designed 200 Gbps IPUs with GPU data bypassing the CPU host on a separate interface compared to other VM network and data traffic.

As with all things supercomputing and cutting-edge AI-based, scalability allows tens of thousands of interconnected GPUs to operate with a "workload bandwidth that is indistinguishable from more expensive off-the-shelf non-blocking network fabrics."

Warren Buffet, one of the world's most influential financial investors, recently appeared at Berkshire Hathaway's annual investors meeting to discuss various topics, one of which was how artificial intelligence will impact the stock market and society as a whole.

The Berkshire Hathaway CEO appeared at the Berkshire Hathaway annual meeting, nicknamed Woodstock for Capitalists, to answer several questions pitched by the company's shareholders. 92-year-old Buffett was joined by Berkshire Hathaway's 99-year-old Vice Chairman Charlie Munger. Buffett was asked by a member of the audience his thoughts on the emergence of artificial intelligence, which he then compared to the creation of the atomic bomb and paraphrased an Albert Einstein quote about the atomic bomb.

Admittedly, Buffett said that artificial intelligence has the power to "change everything in the world" and that it can do "remarkable things". Additionally, Buffett went on to say that when something can do "all kinds of things, I get a little bit worried". However, the world-famous investor led his answer by saying that "there won't be anything in AI that replaces Ajit", Berkshire Hathaway's Vice Chairman of Insurance Operations, then went on to recognize all of its capabilities and the fact that it won't be going away anytime soon.

The impressive nature of OpenAI ChatGPT means it comes with some costs, and these costs aren't Silicon Valley-level speeding fines, according to people familiar with the chatbot's developments.

In a new article posted to The Information, people close to OpenAI's developments of now highly popular ChatGPT, the company spent a staggering $540 million in 2022 training its now extremely impressive AI chatbot. These funds weren't just used to train ChatGPT but also to poach some quality talent from competitor company Google, which is also developing its own AI chatbot called Bard.

The most recent estimated figure for OpenAI's costs is similar to what was estimated by Fortune in January, which put OpenAI's expenses at around $544.5 million for 2022. More specifically, Fortune's estimations outlined costs of $416.45 on computing and data training, $89.31 million on staff, and $38.75 million on unspecified operating costs.

The co-founder of Apple, one of the biggest technology companies on the planet, has thrown some shade at Tesla's AI developments, saying if you wanted to study how AI can kill a human, you should get a Tesla vehicle.

In a recent interview with CNN, Apple co-founder Steve Wozniak discussed various topics around the technology space, and one of those was the developments around artificial intelligence (AI) and its potential dangers. Wozniak was asked during the interview if he ever speaks to Tesla CEO Elon Musk, to which he answered that he's never met Musk in person but does admire some of his technological achievements while distaining others.

Wozniak praised Musk's efforts at pushing large swaths of the population toward electric vehicles but believes the Tesla CEO has made a few empty promises or has fallen short of the capabilities he has promised for his Tesla vehicles. This tune from Wozniak comes with no surprise as the Apple co-founder has long held a critical stance against Tesla's Autopilot feature, saying that the current state of the technology is nowhere near the reality that Elon Musk has promised.

With the rise of artificial intelligence and its exponentially impressive capabilities, many individuals are worried their jobs will be at risk of replacement. Here's an example of why that future may not necessarily be right around the corner.

A video created entirely with AI-powered tools has gone viral on Twitter, and surprisingly, it's a commercial for a beer. The clip called "Synthetic Summer" is a 30-second long video that showcases an AI-generated house party with many people enjoying the advertised beverage. The commercial was created by Helen Power and Chris Boyle from the London-based production company Privateisland.tv.

According to Ars Technica, who wasn't able to reach Power or Boyle for comment before publishing, the video was seemingly created using Runway's Gen-2 AI model, which the publication says is able to generate short video clips based on user prompts in the same way OpenAI's ChatGPT provides responses to text prompts. The 30-second clip was paired with the iconic Shrek song "All-Star" by Smash Mouth, which played over a backyard barbecue scene.

Classic RPG titles such as Skyrim or The Witcher: Wild Hunt are known for being some of the greatest worlds to get lost in, with a large variety of characters to interact and seemingly never-ending adventures to take undertake.

However, gamers reach the end of these games, and with the aforementioned games in particular, they even choose to replay them in a different way for an alternative experience. Unfortunately, there are limitations to these alternative experiences, and one of those is the dialogue players can have with NPCs. Skyrim is one of the most successful games ever made and is known for its iconic dialogue, but what if NPCs were able to have an infinite amount of dialogue and players were able to ask them whatever they wanted?

A very early example of that has been created by modder Art From The Machine, who has taken the underlying language model powering OpenAI's ChatGPT, and combined it with xVASynth for text-to-speech capabilities and Whisper for speech-to-text capabilities. So, what has Art From The Machine achieved? Players are able to speak into their microphones and talk directly to NPCs, who will be able to understand what the player is saying. Whisper converts the player's speech to text which is then fed into ChatGPT's language model that generates a response. That text response is then played through xVASynth, and the NPCs are able to respond appropriately.

75-year-old Dr. Geoffrey Hinton spent more than a decade at Google developing the foundational technology to artificial intelligence, and now "the Godfather of AI" has left the company to sound alarm bells on its potential dangers.

In a new article published in The New York Times, it's explained that Hinton worked at Google for more than 10 years, developing the foundational technology used to create the AI systems we are seeing today. Hinton is renowned as a pioneer of artificial intelligence and gained high respect within his field, but he decided to leave his long career at Google to discuss the dangers of AI and how a part of him regrets the role he played in its AI's initial creation.

In an interview with the NYT, Hinton said, "I console myself with the normal excuse: If I hadn't done it, somebody else would have." Adding, "It is hard to see how you can prevent the bad actors from using it for bad things." Hinton left Google to discuss the potential dangers of AI without any impact from Google and not to criticize the company itself. In fact, Hinton said that Google has acted "very responsibly" with his departure.

Early last month, Samsung staff accidentally leaked an internal source to OpenAI's ChatGPT, sounding alarm bells within the company and outside of it.

There were at least three instances of Samsung employees sharing confidential information with OpenAI's AI-powered chatbot, ChatGPT. The first was a staff member leaking the source code of a confidential database into ChatGPT, which was followed by a request for the AI to check for any errors. Another instance was code optimization being shared with ChatGPT, and the last was a request for ChatGPT to convert an internal Samsung video of a meeting into minutes.

Now, Bloomberg News has read a new memo issued to staff notifying them of a new policy change. According to the memo, Samsung is terribly concerned about the rise of AI-powered tools and how it may affect intellectual property. In an effort to reduce any leaks of confidential data that may or may not impact the company, a widespread ban has been placed on any AI-generative tools on employee-issued devices. Samsung staff are now prohibited from having any AI-generative tools or applications on company-owned computers, tablets, phones, and on any of Samsung's internal networks.

Scientists announced on Monday that they created a way for an artificial intelligence system to transcribe what people are thinking by feeding it scans of the individual's brain activity.

The new system was developed by researchers at The University of Texas in Austin, and according to the study published in the journal Nature Neuroscience, the goal behind this new technology is to provide assistance to people that are mentally conscious but are unable to physically speak, such as people that have suffered from strokes. The study was led by Jerry Tang, a doctoral student in computer science, and Alex Huth, an assistant professor of neuroscience and computer science at UT Austin.

The researchers explain that the AI model was trained on people that participated in several sessions of magnetic resonance imaging (fMRI). These long hours of brain activity recording were then fed into GPT-1, a predecessor language model that was later developed into GPT-4, the language model powering the popular AI chatbot, ChatGPT. Scientists then trained the model to predict how each person's brain would respond to hearing speech, such as a short story. These stories were listened to by participants while they were in the fMRI sessions.

NVIDIA has unveiled a new piece of software, NeMo Guardrails, that will ensure chatbots driven by large language models (LLMs) - such as ChatGPT (the engine of Microsoft's Bing AI) - stay on track in various ways.

The biggest problem the software sets out to resolve is what's known as 'hallucinations,' occasions when the chatbot goes awry and makes an inaccurate or even absurd statement.

These are the kind of incidents that were reported in the early usage of Bing AI and Google's Bard. They're often embarrassing episodes, frankly, which erode trust in the chatbot for obvious reasons.

OpenAI's ChatGPT has pioneered AI-powered chatbots such as Microsoft's Bing Chat or Google's Bard. But how much does it cost developers of these chatbots to keep them up and running?

A new report posted on The Information cites Dylan Patel, chief analyst at semiconductor research firm SemiAnalysis, who said that OpenAI could be paying as much as $700,000 a day to keep ChatGPT servers up. So, why does ChatGPT cost so much to run? It's relatively simple. ChatGPT requires a large amount of power to analyze its database and create an appropriate response for a prompt. Patel spoke to Insider and said that his initial estimate was based on OpenAI's GPT-3.5 model, which is far less powerful than OpenAI's most-recent model, GPT-4.

Patel says that GPT-4 would cost the company much more money simply because that language model has many times more parameters. Furthermore, speaking to Forbes, Patel and Afzal Ahmad, another analyst from SemiAnalysis, said that it would have likely cost tens of millions of dollars to train ChatGPT's underlying language models, but that cost is nothing compared to operational expenses or inference costs. The analysts said that ChatGPT's inference costs "exceed the training costs on a weekly basis".

The European Consumer Organisation (BEUC) is calling for consumer protection bodies in Europe to investigate ChatGPT and similar chatbot technologies to ascertain how much risk they might represent to the public.

Reuters reports that the BEUC - an umbrella group covering consumer protection organizations across 32 countries - is stepping up to challenge whether such chatbots could be problematic in terms of their influence on youngsters in particular.

The worry is that the responses of the AI to queries may appear authoritative and true - particularly to younger consumers and children - but cannot be relied upon, and indeed are often factually incorrect to some degree (perhaps even a large one).

An AI-generated song that attracted millions of plays has been removed after the Universal Music Group (UMG) discovered its popularity and the artists that were on track.

The song called 'Heart On My Sleeve' was released by an anonymous TikTok user called Ghostwriter977, and it featured vocals from Aubrey "Drake" Graham and Abel Makkonen "The Weeknd" Tesfaye. But these vocals didn't come from the artists themselves, they were created using artificial intelligence-powered tools that were fed voice samples of the artists to create imitated voices that are uncannily similar. At the moment, it remains unclear if the entire song was created with AI tools or just the voices of the artists.

Regardless of the intricacies of how the song was made, it generated more than 1,000,000 streams on Spotify, and Ghostwriter977's TikTok video of the song was viewed more than 15 million times. Additionally, a YouTube video of the song gained more than 275,000 views, and its creator commented on it, saying, "this is just the beginning". Overall the response to the song was very positive, with some people saying it was the best Drake song that has been released this year. Others recognized the quality of the song and immediately commented on how good AI is getting at sampling an artist's vocals.

Google staff members weren't too impressed with the state of Bard when asked to test the AI before its launch, and indeed warned the company that the AI wasn't ready to meet its public.

This is according to a Bloomberg report, which relayed information from some 18 current and former employees of Google, as well as from internal documents - and the overall vibe is of stark warnings that Bard needed more polish.

Bard was labeled as 'cringeworthy' and a 'pathological liar' alongside being called 'worse than useless' in an internal message to thousands of staffers, many of whom agreed. The comment about being useless was followed by a plea not to launch Bard, we're told.

According to reports, Microsoft is working on its own AI hardware, code-named Athena, that will be built using TSMC's 5nm process. Microsoft has been working on the project since 2019 (as per The Information) to directly respond to the surging costs of running AI hardware.

The report indicates that a small group of Microsoft and OpenAI employees is already testing the chips. It aims to perform better than the hardware it currently buys from vendors like NVIDIA, not to mention save them a lot of money, with the AI boom showing no signs of slowing down.

Microsoft has invested billions with Open AI, the makers of ChatGPT, and is in the midst of an AI arms race when implementing models into its product suite. These AI chips could be ready and widely available as early as next year, making them a potential revenue stream for Microsoft and Azure.

A mother in Arizona has received a terrifying phone call from what seemed to be a loved one about her daughter's kidnapping.

Arizona resident Jennifer DeStefano told WKYT that she received a phone call from a random phone number while she was at one of her daughter's dance studios. DeStefano said she almost let the phone call ring through but was concerned that her other 15-year-old daughter, that was out on a skiing trip, may have been in trouble. DeStefano answered the call and immediately heard her sobbing 15-year-old daughter's voice crying out, "Mom!".

Panicked, DeStefano quickly asked what was wrong, and her daughter's voice said, "Mom, I messed up," which was followed by more sobbing and crying. Before another question could be asked by DeStefano, she recounted a man yelling in the background of the call, "Put your head back, lie down". This unknown man got on the phone and demanded that DeStefano pay him $1 million dollar for the safe return of her daughter while also saying that if she were to call the police or anybody else that he was going to pump her daughter full of drugs, violate her and then drop her off in Mexico.