Cloud Hosting

Some researchers think that AI could eventually achieve general intelligence, matching and even exceeding humans on most tasks.Credit: Charles Taylor/Alamy

Will an artificial intelligence (AI) superintelligence appear suddenly, or will scientists see it coming, and have a chance to warn the world? Thats a question that has received a lot of attention recently, with the rise of large language models, such as ChatGPT, which have achieved vast new abilities as their size has grown. Some findings point to emergence, a phenomenon in which AI models gain intelligence in a sharp and unpredictable way. But a recent study calls these cases mirages artefacts arising from how the systems are tested and suggests that innovative abilities instead build more gradually.

I think they did a good job of saying nothing magical has happened, says Deborah Raji, a computer scientist at the Mozilla Foundation who studies the auditing of artificial intelligence. Its a really good, solid, measurement-based critique.

The work was presented last week at the NeurIPS machine-learning conference in New Orleans.

Large language models are typically trained using huge amounts of text, or other information, whch they use to generate realistic answers by predicting what comes next. Even without explicit training, they manage to translate language, solve mathematical problems and write poetry or computer code. The bigger the model is some have more than a hundred billion tunable parameters the better it performs. Some researchers suspect that these tools will eventually achieve artificial general intelligence (AGI), matching and even exceeding humans on most tasks.

ChatGPT broke the Turing test the race is on for new ways to assess AI

The new research tested claims of emergence in several ways. In one approach, the scientists compared the abilities of four sizes of OpenAIs GPT-3 model to add up four-digit numbers. Looking at absolute accuracy, performance differed between the third and fourth size of model from nearly 0% to nearly 100%. But this trend is less extreme if the number of correctly predicted digits in the answer is considered instead. The researchers also found that they could also dampen the curve by giving the models many more test questions in this case the smaller models answer correctly some of the time.

Next, the researchers looked at the performance of Googles LaMDA language model on several tasks. The ones for which it showed a sudden jump in apparent intelligence, such as detecting irony or translating proverbs, were often multiple-choice tasks, with answers scored discretely as right or wrong. When, instead, the researchers examined the probabilities that the models placed on each answer a continuous metric signs of emergence disappeared.

Finally, the researchers turned to computer vision, a field in which there are fewer claims of emergence. They trained models to compress and then reconstruct images. By merely setting a strict threshold for correctness, they could induce apparent emergence. They were creative in the way that they designed their investigation, says Yejin Choi, a computer scientist at the University of Washington in Seattle who studies AI and common sense.

Study co-author Sanmi Koyejo, a computer scientist at Stanford University in Palo Alto, California, says that it wasnt unreasonable for people to accept the idea of emergence, given that some systems exhibit abrupt phase changes. He also notes that the study cant completely rule it out in large language models let alone in future systems but adds that "scientific study to date strongly suggests most aspects of language models are indeed predictable.

Raji is happy to see the community pay more attention to benchmarking, rather than to developing neural-network architectures. Shed like researchers to go even further and ask how well the tasks relate to real-world deployment. For example, does acing the LSAT exam for aspiring lawyers, as GPT-4 has done, mean that a model can act as a paralegal?

The work also has implications for AI safety and policy. The AGI crowd has been leveraging the emerging-capabilities claim, Raji says. Unwarranted fear could lead to stifling regulations or divert attention from more pressing risks. The models are making improvements, and those improvements are useful, she says. But theyre not approaching consciousness yet.

See the original post here:

Will superintelligent AI sneak up on us? New study offers reassurance - Nature.com

We lost Twitter and got X. We tried out Bluesky and Mastodon (well, some of us did). We fretted about AI bots and teen mental health. We cocooned in private chats and scrolled endlessly as we did in years past. For social media users, 2023 was a year of beginnings and endings, with some soul-searching in between.

Heres a look back some of the biggest stories in social media in 2023 and what to watch for next year:

A little more than a year ago, Elon Musk walked into Twitter s San Francisco headquarters, fired its CEO and other top executives and began transforming the social media platform into whats now known as X.

Musk revealed the X logo in July. It quickly replaced Twitters name and its whimsical blue bird icon, online and on the companys San Francisco headquarters.

And soon we shall bid adieu to the twitter brand and, gradually, all the birds, Musk posted on the site.

Because of its public nature and because it attracted public figures, journalists and other high-profile users, Twitter always had an outsized influence on popular culture but that influence seems to be waning.

It had a lot of problems even before Musk took it over, but it was beloved brand with a clear role in the social media landscape, said Jasmine Enberg, a social media analyst at Insider Intelligence. There are still moments of Twitter magic on the platform, like when journalists took the platform to post real-time updates about the OpenAI drama, and the smaller communities on the platform remain important to many users. But the Twitter of the past 17 years is largely gone, and Xs reason for existence is murky.

Since Musks takeover, X has been bombarded by allegations of misinformation and racism, endured significant advertising losses and suffered declines in usage. It didnt help when Musk went on an expletive-ridden rant in an on-stage interview about companies that had halted spending on X. Musk asserted that advertisers that pulled out were engaging in blackmail and, using a profanity, essentially told them to get lost.

Continuing the trend of welcoming back users who had been banned by the former Twitter for hate speech or spreading misinformation, in December, Musk restored the X account of conspiracy theorist Alex Jones, pointing to an unscientific poll he posted to his followers that came out in favor of the Infowars host who repeatedly called the 2012 Sandy Hook school shooting a hoax.

LGBTQ and other organizations supporting marginalized groups, meanwhile, have been raising alarms about X becoming less safe. In April, for instance, it quietly removed a policy against the targeted misgendering or deadnaming of transgender individuals. In June, the advocacy group GLAAD called it the most dangerous platform for LGBTQ people.

GLSEN, an LGBTQ education group, announced in December that it was leaving X, joining other groups such as the suicide prevention nonprofit Trevor Project, saying that Musks changes have birthed a new platform that enables its users to harass and target the LGBTQ+ community without restriction or discipline.

Musks ambitions for X include transforming the platform into an everything app like Chinas WeChat, for instance. The problem? Its not clear if U.S. and Western audiences are keen on the idea. And Musk himself has been pretty vague on the specifics.

While X contends with an identity crisis, some users began looking for a replacement. Mastodon was one contender, along with Bluesky, which actually grew out of Twitter a pet project of former CEO Jack Dorsey, who still sits on its board of directors.

When tens of thousands of people, many of them fed-up Twitter users, began signing up for the (still) invite-only Bluesky in the spring, the app had less than 10 people working on it, said CEO Jay Graber recently.

This meant scrambling to keep everything working, keeping people online, scrambling to add features that we had on the roadmap, she said. For weeks, the work was simply scaling ensuring that the systems could handle the influx.

We had one person on the app for a while, which was very funny, and there were memes about Paul versus all of Twitters engineers, she recalled. I dont think we hired a second app developer until after the crazy growth spurt.

Seeing an opportunity to lure in disgruntled Twitter users, Facebook parent Meta launched its own rival, Threads, in July. It soared to popularity as tens of millions began signing up though keeping people on has been a bit of a challenge. Then, in December, Meta CEO Mark Zuckerberg announced in a surprise move that the company was testing interoperability the idea championed by Mastodon, Bluesky and other decentralized social networks that people should be able to use their accounts on different platforms kind of like your email address or phone number.

Starting a test where posts from Threads accounts will be available on Mastodon and other services that use the ActivityPub protocol, Zuckerberg posted on Threads in December. Making Threads interoperable will give people more choice over how they interact and it will help content reach more people. Im pretty optimistic about this.

Social medias impact on childrens mental health hurtled toward a reckoning this year, with the U.S. surgeon general warning in May that there is not enough evidence to show that social media is safe for children and teens and calling on tech companies, parents and caregivers to take immediate action to protect kids now.

Were asking parents to manage a technology thats rapidly evolving that fundamentally changes how their kids think about themselves, how they build friendships, how they experience the world and technology, by the way, that prior generations never had to manage, Dr. Vivek Murthy told The Associated Press. And were putting all of that on the shoulders of parents, which is just simply not fair.

In October, dozens of U.S. states sued Meta for harming young people and contributing to the youth mental health crisis by knowingly and deliberately designing features on Instagram and Facebook that addict children to its platforms.

In November, Arturo Bjar, a former engineering director at Meta, testified before a Senate subcommittee about social media and the teen mental health crisis, hoping to shed light on how Meta executives, including Zuckerberg, knew about the harms Instagram was causing but chose not to make meaningful changes to address them.

The testimony came amid a bipartisan push in Congress to adopt regulations aimed at protecting children online. In December, the Federal Trade Commission proposed sweeping changes to a decades-old law that regulates how online companies can track and advertise to children, including turning off targeted ads to kids under 13 by default and limiting push notifications.

Your AI friends have arrived but chatbots are just the beginning. Standing in a courtyard at his companys Menlo Park, California headquarters, Zuckerberg said this fall that Meta is focused on building the future of human connection and painted a near-future where people interact with hologram versions of their friends or coworkers and with AI bots built to assist them. The company unveiled an army of AI bots with celebrities such as Snoop Dogg and Paris Hilton lending their faces to play them that social media users can interact with.

Next year, AI will be integrated into virtually every corner of the platforms, Enberg said.

Social apps will use AI to drive usage, ad performance and revenues, subscription sign ups, and commerce activity. AI will deepen both users and advertisers reliance and relationship with social media, but its implementation wont be entirely smooth sailing as consumer and regulatory scrutiny will intensify, she added.

The analyst also sees subscriptions as an increasingly attractive revenue stream for some platforms. Inspired by Musks X, subscriptions started as a way to diversify or boost revenues as social ad businesses took a hit, but they have persisted and expanded even as the social ad market has steadied itself.

With major elections coming up in the U.S. and India among other countries, AIs and social medias role in misinformation will continue to be front and center for social media watchers.

Were not prepared for this, A.J. Nash, vice president of intelligence at the cybersecurity firm ZeroFox, told the AP in May. To me, the big leap forward is the audio and video capabilities that have emerged. When you can do that on a large scale, and distribute it on social platforms, well, its going to have a major impact.

Excerpt from:

The year of social media soul-searching: Twitter dies, X and Threads are born and AI gets personal - The Associated Press

Bill Gates' first grandchild was born in 2023, so the year will forever be special to him, he says. It was also the year that artificial intelligence went mainstream, spurred by the arrival of ChatGPT. And that got Gates thinking about how the world his granddaughter is coming into will change in a positive way because of AI.

The co-founder of Microsoft and a guiding figure of the PC era across several decades, Gates knows a thing or two about technological revolutions. He sees 2024 as a monumental year for artificial intelligence, with the technology becoming especially important in global health, where Gates and his namesake foundation have been working for decades.

"We now have a better sense of what types of jobs AI will be able to do by itself and which ones it will serve as a copilot for," Gates wrote in a lengthy post on his GatesNote blog this week. "And it's clearer than ever how AI can be used to improve access to education, mental health, and more. It motivates me to make sure this technology helps reduce -- and doesn't contribute to -- the awful inequities we see around the world."

It's been quite a year for AI, and more specifically, generative AI. Gen AI goes a step further than other AI methods. It can create new materials, such as text, images, speech or video, based on its own understanding of the patterns it recognizes in data.

Gen AI became known thanks to the launch of OpenAI's ChatGPT in late 2022, although smart home controls and AI-powered virtual assistants such as Alexa had already made inroads into homes and popular culture.

ChatGPT, the frontrunner in the onslaught of generative AI tools released over the last year, allows anyone with a smartphone or a laptop to use AI for generating information or images. These tools been trained on huge swaths of data that allow it to come up with original responses to our queries -- with varying degrees of success. More than 100 million people use ChatGPT each week, OpenAI chief executive Sam Altman said in November. Microsoft is a significant investor in OpenAI.

Other companies aren't ceding territory to Microsoft. In early December, for instance, Google began updating its Bard AI chatbot with a new AI model called Gemini that provides improved text-based chat abilities. Tech companies are continuing to add Gen AI abilities into programs and devices of all kinds, from search engines to smart phones.

In 2023, investors pourednearly $10 billioninto gen AI startups, more than double the $4.4 billion invested the year before, according toGlobalData.

But even as Gen AI explodes in popularity, many users are still cautious. In addition to concerns that AI could replace human employees, many worry about it putting forth inaccurate information. Dictionary.com selected the AI term "hallucinate," describing what happens when AI produces false information, as its word of the year.

Gates thinks mainstream integration of AI is coming soon.

"If I had to make a prediction, in high-income countries like the United States, I would guess that we are 18-24 months away from significant levels of AI use by the general population," he wrote.

But he also sees 2024 as a turning point.

Since stepping down as Microsoft CEO in 2000, Gates has focused on philanthropy, founding the Bill & Melinda Gates Foundation with his now-former wife. It's in areas related to his foundation's work in global health where Gates sees AI becoming helpful in 2024.

Fighting antibiotic resistance: He cites an AI-powered tool under development at the Aurum Institute in Ghana that helps health workers prescribe antibiotics without contributing to antimicrobial resistance, where pathogens learn how to get past antibiotic defenses. The tool can comb through all the available information about antimicrobial resistance and suggest the best drug plan for a patient.

High-risk pregnancy help: A woman dies in childbirth every two minutes, Gates says. He's hopeful that AI can combat this horrifying statistic. AI-powered ultrasounds can help identify pregnancy risks, and the Gates foundation is working to fund that process. Also, AI researchers at ARMMAN, an India-based nonprofit organization, are working on a large language model the technology that underlies ChatGPT and other AI chatbots that can help health workers treating high-risk pregnancies.

HIV risk assessment: Many people aren't comfortable talking to a doctor about their sexual history, but that can be vital for assessing risk for diseases like HIV. Gates is excited about a South African chatbot called Your Choice, being developed by Sophie Pascoe of Wits Health Consortium. The chatbot acts as a nonjudgmental counselor that can provide round-the-clock advice, especially for vulnerable populations.

Quick access to medical records: While people in rich countries may have their medical records easily available, in other countries, many people don't have a documented medical history, Gates says. This can hinder their medical treatment because their doctors need to know about allergies, past health issues and more. A Pakistani team is working on a voice-enabled mobile app that could make this easier, asking a series of questions and filling out a patient's medical record with the answers.

Gates also sees AI assisting in education, calling AI education tools "mindblowing," as they are tailored to individual learners, and says they will "only get better." He's excited about how the technology can be localized to students in many different countries and cultural contexts.

Not everything on Gates' mind is AI-related. He's concerned about climate change, saying he's "blown away by the passion from young climate activists," and hopeful that 2024 will see more investment in innovations that will help those who are most affected by the climate crisis.

And he even plunges into the debate over nuclear energy. Gates notes that high-profile disasters such as Chernobyl in the 1980s and Three Mile Island in the late 1970s have spotlighted the risks, but over the past year, he's seen a shift towards acceptance. He sees the once-bogeyman of the energy world as necessary to meet the world's growing need for energy while eliminating carbon emissions.

A New York Times in early December noted that Gates was "long skeptical" of what AI could do. That changed in August 2022, when he saw a demonstration of OpenAI's GPT-4, the large language model underlying ChatGPT. That sold Gates on the concept, and he helped Microsoft "move aggressively to capitalize on generative AI."

Although Gates left Microsoft's's board in 2020, he's still an adviser to its CEO Satya Nadella. Microsoft has plunged full-bore into the AI world, The company invested heavily in OpenAI, the creator of ChatGPT, earlier this year. And it's been adding the technology across its online services, including its Bing search engine.

The company also reimagined Windows 11 with the addition of Microsoft Copilot, which puts AI assistance always available on the Windows 11 desktop taskbar. Microsoft vice president Yusuf Mehdi calls it the most significant update to the operating system so far, and it works across multiple apps and mobile phones.

Gates' year-end letter compares the rise of AI to that of the internet, email and search engines, noting that it wasn't long ago when many people were unfamiliar with these things, and now they are part of our daily lives. Gates sees the same kind of sea change coming with AI.

But he admits that it won't be smooth, giving an example from his own life.

"I thought I would use AI tools for the foundation's strategy reviews this year, which require reading hundreds of pages of briefing materials that an AI could accurately summarize for me," Gates says.

But that didn't happen.

"Old habits are hard to break, and I ended up preparing for [the reviews] the same way I always do," he writes.

Editors' note: CNET is using an AI engine to help create some stories. For more, see our AI policy.

Read this article:

Why Bill Gates Says AI Will Supercharge Medical Innovations - CNET

Artificial intelligence has been inescapable this year. After OpenAI released ChatGPT some 13 months ago, attention turned to how such tools will disrupt careers and industriesand eager venture capitalists poured billions into AI startups that might do the disrupting.

But VCs themselves could get disrupted, according to billionaire investor Chamath Palihapitiya, a former Facebook executive and the CEO of VC firm Social Capital.

We talk about AI as a big disruptor to the big companies and this and that, but AI may be the biggest disruptor to VC in the end, Palihapitiya said on the All-In Podcast this week.

A world where AI proliferates, he said, is positive for founders, who will be able to own more of their companies rather than give away too much equity to VCs.

In the past, he said, a tech startup with $2 million in seed funding might hire seven people and have enough capital to survive for a year and a half, after which it hopefully gained enough traction so that investors would pony up $10 million or $15 million in Series A funding. The downside, of course, is that in exchange for capital, VCs want equity in the company.

But AI tools give founders more leverage, Palihapitiya said, mentioning GitHub Copilot, which makes creating and fixing code much easier. Startups can now hire programmers, perhaps in other countries with lower pay rates, to use such tools to get more done faster, he noted.

The upshot is that, today, a tech startup with the same amount of seed funding might have a three- or four-person team and survive on that $2 million for four years rather than a year a half. Founders could then end up owning 80% of their company with the potential to exit for $50 million or $100 million, and theyve made more money than in a traditional outcome he said.

Its only a matter of time, Palihapitiya added, until they can put two and two together in an Excel spreadsheet to figure out that owning 50% of a $100 million company is greater than owning 18% of some other company when youre massively diluted, or 8% or whatever.

Jason Calacanis, an angel investor, responded that now, instead of founders of a particular cohort competing on who can raise the most money at the highest valuation, hes seen them shifting to, how do I get to profitability and how do I own as much of my company as possible?

Palihapitiya became the face of the SPAC boom-and-bust a few years back due to his involvement with special purpose acquisition companiesshell corporations listed on a stock exchange that acquire a private company, thereby making it public sans the rigors of the IPO process.

This isnt the first time he has mulled the role of VCs in an AI-altered world.

It seems pretty reasonable and logical, he said last month on the podcast, that AI productivity gains will lead to tens or hundreds of millions of startups made up of only one or two people.

Theres a lot of sort of financial engineering that kind of goes away in that world, he said. I think the job of the venture capitalist changes really profoundly. I think theres a reasonable case to make that it doesnt exist.

Excerpt from:

Chamath Palihapitiya says venture capitalists also face disruption from AIand startup founders stand to benefit - Fortune