Newswise — ALBUQUERQUE, N.M. — Social media, cameras, sensors and more generate huge amounts of data that can overwhelm analysts sifting through it all for meaningful, actionable information to provide decision-makers such as political leaders and field commanders responding to security threats.

Sandia National Laboratories researchers are working to lessen that burden by developing the science to gather insights from data in nearly real time.

“The amount of data produced by sensors and social media is booming — every day there’s about 2.5 quintillion (or 2.5 billion billion) bytes of data generated,” said Tian Ma, a Sandia computer scientist and project co-lead. “About 90% of all data has been generated in the last two years — there’s more data than we have people to analyze. Intelligence communities are basically overwhelmed, and the problem is that you end up with a lot of data sitting on disks that could get overlooked.”

Sandia researchers worked with students at the University of Illinois Urbana-Champaign, an Academic Alliance partner, to develop analytical and decision-making algorithms for streaming data sources and integrated them into a nearly real-time distributed data processing framework using big data tools and computing resources at Sandia. The framework takes disparate data from multiple sources and generates usable information that can be acted on in nearly real time.

To test the framework, the researchers and the students used Chicago traffic data such as images, integrated sensors, tweets and streaming text to successfully measure traffic congestion and suggest faster driving routes around it for a Chicago commuter. The research team selected the Chicago traffic example because the data inputted has similar characteristics to data typically observed for national security purposes, said Rudy Garcia, a Sandia computer scientist and project co-lead.

Drowning in data

“We create data without even thinking about it,” said Laura Patrizi, a Sandia computer scientist and research team member, during a talk at the 2019 United States Geospatial Intelligence Foundation’s GEOINT Symposium. “When we walk around with our phone in our pocket or tweet about horrible traffic, our phone is tracking our location and can attach a geolocation to our tweet.”

To harness this data avalanche, analysts typically use big data tools and machine learning algorithms to find and highlight significant information, but the process runs on recorded data, Ma said.

“We wanted to see what can be analyzed with real-time data from multiple data sources, not what can be learned from mining historical data,” Ma said. “Actionable intelligence is the next level of data analysis where analysis is put into use for near-real-time decision-making. Success on this research will have a strong impact to many time-critical national security applications.”

Building a data processing framework

The team stacked distributed technologies into a series of data processing pipelines that ingest, curate and index the data. The scientists wrangling the data specified how the pipelines should acquire and clean the data.

“Each type of data we ingest has its own data schema and format,” Garcia said. “In order for the data to be useful, it has to be curated first so it can be easily discovered for an event.”

Hortonworks Data Platform, running on Sandia’s computers, was used as the software infrastructure for the data processing and analytic pipelines. Within Hortonworks, the team developed and integrated Apache Storm topologies for each data pipeline. The curated data was then stored in Apache Solr, an enterprise search engine and database. PyTorch and Lucidwork’s Banana were used for vehicle object detection and data visualization.

Finding the right data

“Bringing in large amounts of data is difficult, but it’s even more challenging to find the information you’re really looking for,” Garcia said. “For example, during the project we would see tweets that say something like ‘Air traffic control has kept us on the ground for the last hour at Midway.’ Traffic is in the tweet, but it’s not relevant to freeway traffic.”

To determine the level of traffic congestion on a Chicago freeway, ideally the tool could use a variety of data types, including a traffic camera showing flow in both directions, geolocated tweets about accidents, road sensors measuring average speed, satellite imagery of the areas and traffic signs estimating current travel times between mileposts, said Forest Danford, a Sandia computer scientist and research team member.

“However, we also get plenty of bad data like a web camera image that’s hard to read, and it is rare that we end up with many different data types that are very tightly co-located in time and space,” Danford said. “We needed a mechanism to learn on the 90 million-plus events (related to Chicago traffic) we’ve observed to be able to make decisions based on incomplete or imperfect information.”

The team added a traffic congestion classifier by training merged computer systems modeled on the human brain on features extracted from labeled images and tweets, and other events that corresponded to the data in time and space. The trained classifier was able to generate predictions on traffic congestion based on operational data at any given time point and location, Danford said.

Professors Minh Do and Ramavarapu Sreenivas and their students at UIUC worked on real-time object and image recognition with web-camera imaging and developed robust route planning processes based off the various data sources.

“Developing cogent science for actionable intelligence requires us to grapple with information-based dynamics,” Sreenivas said. “The holy grail here is to solve the specification problem. We need to know what we want before we build something that gets us what we want. This is a lot harder than it looks, and this project is the first step in understanding exactly what we would like to have.”

Moving forward, the Sandia team is transferring the architecture, analytics and lessons learned in Chicago to other government projects and will continue to investigate analytic tools, make improvements to the Labs’ object recognition model and work to generate meaningful, actionable intelligence.

“We’re trying to make data discoverable, accessible and usable,” Garcia said. “And if we can do that through these big data architectures, then I think we’re helping.”

The post Wrangling big data into real-time, actionable intelligence appeared first on Artificial Intelligence.

In his 1990 book The Age of Intelligent Machines, the American computer scientist and futurist Ray Kurzweil made an astonishing prediction. Working at the Massachusetts Institute of Technology (MIT) throughout the 1970s and 1980s and having seen firsthand the remarkable advances in artificial intelligence pioneered there by Marvin Minsky and others, he forecast that a computer would pass the Turing test – the test of a machine’s ability to match or be indistinguishable from human intelligence – between 2020 and 2050.

Kurzweil, now Google’s head of artificial intelligence, or AI (an acronym with which we’ve all now become familiar), has subsequently refined his claim. He now says this event will happen by 2029. What’s more, in 2045 we will witness what he calls “the singularity” – the point at which human and artificial intelligences merge, leading to exponential advances in technology and human capabilities.

Exciting stuff. Or is it? While Kurzweil is famously optimistic about the effect AI will have on human lives, others aren’t so sure. Part of this stems from the fear, inculcated by a thousand sci-fi movies, that “the robots will take over” – either rendering humans functionally useless or worse, becoming our masters in a dystopian role reversal.

The truth may lie somewhere in between. But what’s increasingly clear is that AI is advancing at a rapid pace. Already, it is posing profound questions about the future of work, of society and the very nature of what it means to be human.

We can get a sense for this from current innovations. Whether it’s self-driving vehicles, devices like Amazon’s Echo that can “understand” human language, or the intelligent crunching of vast medical datasets to diagnose disease more accurately, we’re moving towards a place where all manner of tasks are automated and human error – or perhaps human judgment – is obviated.

A recent report by the global consultancy McKinsey estimated that almost half (49 per cent) of the activities people are paid nearly $16 trillion in wages to do in the global economy have the potential to be automated by adapting currently demonstrated technologies.

This month I had the chance to discuss some of these implications before a House of Lords select committee in the UK’s Parliament. One issue the House of Lords committee is considering is whether to recommend the appointment of a new minister for AI to provide a coordinated response to these shifting sands across different government departments.

In this, the UK would be playing catch-up to the UAE, which, in a world-first, recently appointed 27-year-old Omar bin Sultan Al Olama to this role. Speaking to The National recently, Mr Al Olama set out his positive, practical vision for AI, saying it could offer humanity a “quick win” in helping to tackle climate change and other pressing problems.

While the UAE might be famous abroad for its glittering, futuristic cities of Dubai and Abu Dhabi, it is also situated in one of the world’s most ecologically sensitive regions and has an acute economic need to move away from fossil fuels as the main source of its wealth – so government-led nurturing of AI’s potential for sustainable ends makes a lot of sense.

Nor is the UAE alone in wanting to reap the economic benefits of AI. Indeed, across the world, it’s no exaggeration to say that something of an AI arms race has begun. China recently announced its intention to dominate the sector, creating a $150 billion industry by 2030, in direct competition with the US. China’s output of academic papers on artificial intelligence overtook the 28 EU countries combined for the first time last year.

Of course, we all hope that technological advances will continue to happen and will continue, on balance, to benefit humanity. But there’s little doubt there will be some unintended consequences, some of which are already manifesting themselves. One thing I’m particularly concerned about is the impact on job security. Sitting next to me in the House of Lords committee meeting was Olly Buston, the chief executive of the think-tank Future Advocacy, which recently published a report estimating between 22 per cent and 39 per cent of jobs in the UK are at high risk from automation by the early 2030s. With an average of about 30 per cent across the country, that represents more than 10 million livelihoods.

In London, where I work, there have been controversies recently over the way new technology-driven companies like Uber, a taxi service, and Deliveroo, a takeaway food delivery company, treat their workers. Because their businesses are built around advanced data services and smartphone apps, they have access to vast numbers of potential customers and willing recruits at the push of a button. But they have chosen, somewhat cynically, to class the people who work for them as self-employed rather than as employees. In the UK, this means they have no automatic right to sick pay, holiday or pension contributions and no opportunity for career progression.

Replicated at scale across an economy, you can imagine the effect this could have. While many workers will see their job security and welfare safety net vanish, those in control of the technology stand to benefit enormously. The question then becomes: how can we change the social contract so that we don’t just see runaway inequality and wealth polarisation? How can we make sure the increases in productivity and value-generation AI promises benefit all of society, instead of forcing millions into a precarious, hand-to-mouth existence that leaves little room for personal flourishing?

Recently, the Institute for Global Prosperity, of which I am director, put forward a radical new proposal that could help address one aspect of this challenge. Known as universal basic services, this would see many of the essentials of 21st century life – including housing, food, transport and information technologies – provided free at the point of need. This is a familiar concept in the UK, where our National Health Service has been providing needs-based healthcare for all, for the past 70 years. We have calculated that an extension of provision into these other areas could be afforded at a cost equivalent to about 2.3 per cent of the UK’s GDP. The practical effect of this would be to dramatically reduce the basic cost of living for most people, giving them greater freedom over their work and leisure choices.

We’re not saying this would be a panacea. But compared to other ideas like a universal basic income – a flat payment to all citizens – it’s far more affordable and could be one measure that helps to alleviate the worst impacts of inequality brought about by an increasingly tech-driven world.

This needs to go hand-in-hand with a much more detailed analysis of how we can upskill our populations through education to make the most of AI’s positive potential. Kurzweil said recently that AI will enhance us, not replace us. His predictions have often been right – but that doesn’t mean we shouldn’t start planning now for this radically changed future.

The post Artificial intelligence will enhance us, not replace us appeared first on Artificial Intelligence.

OAKLAND, Calif. (KTVU) – Just like a gun, a computer can do good or evil depending on the user. But consider an artificially intelligent gun or computer that can think for itself. Thursday, we looked at the dicey ethical issues of autonomous computers, machines and robots.

In Isaac Asimov’s sci-fi classic: I, Robot, robots were always required to obey three laws. A robot must obey human orders except where such orders would conflict with the First Law. A robot must protect its own existence so it does not conflict with the First or Second Laws. But even that didn’t work out.

Though they do our bidding now, eventually artificially intelligent machines and robots could become self-aware life forms unto themselves, as they learn more and more.

Some theoretical physicist, such as Stephen Hawking and tech guru Elon Musk, suggest that this could become humankind’s undoing.

“I mean, with artificial intelligence, we are summoning the demon,” said Musk. “Conceivably be destroyed by it,” said Hawking.

Unlike humans, robots are tireless, relentless, don’t call in sick and are driven by their programming. Because of that, they can do work that is boring, tedious and tiring to humans – faster and more accurately and without human passion or prejudice.

But when they can think for themselves will they develop biases and perhaps, even become anti-human?

Irina Raicu is the University of Santa Clara’s Internet Ethics Program Director. “It’s the first time that we’re really talking about whether the machines will really do the kinds of things that make us human. Will they care for people, think through the implications of their actions, revolt if they are given bad orders or something? They might recreate the biases present in society, rather than correct them or they may just reach really bad decisions,” said Raicu.

For the moment, however, Raicu sees artificial intelligence ethics as a problem far down the technological road.

“The kind of ethical reasoning that you’re describing is done by people and will have to continue to be done by people for the foreseeable future,” he said.

Stanford Computer Scientist Michael Genesereth agrees. “I’m pretty sure that humans today could not live without their machines. It seems they have become essential. But. I am sure, certainly, about one thing: our machines cannot live without humans, at least not for a very long time. We are nowhere near where they will subsist entirely on their own.” said Genesereth.

Genesereth sites the case where the autonomous car – the self-driving car – is faced with deciding which way to swerve. One way this six people and the other way one person.

“And that kind of reasoning is not programmed and not taught into machines, although there are people who are trying because there are decisions that have to be made that have an ethical component,” said Raicu. “Invariably, so far, human beings have a better sense of how to respond. The machines are nowhere near that point.”

But, most observers agree, we will reach a point one day where artificial intelligence will take many of our jobs and, if we’re not careful, perhaps our humanity.

The post Artificial intelligence: robots with ethics? appeared first on Artificial Intelligence.

Computer scientist and celebrated futurist Ray Kurzweil says artificial intelligence will match human intelligence by 2029, and that by 2045, it will have multiplied the human biological machine intelligence of our civilization a billion times.

Predictions like this abound amid the hype surrounding AI. Real understanding however, is less common. Many enterprises are unclear about what constitutes AI, where it can be applied, and how to prioritize its use cases within the organization. In a survey commissioned by Infosys on the state of AI adoption, half of the 1,600 respondents said that not knowing where AI could assist was one of their biggest challenges.

A good understanding of the technologies under the umbrella of artificial intelligence is key. Many of these technology pieces have seen rapid and impressive evolution in the recent past, so much so that it does not allow banking providers the luxury of waiting until it matures. Here are the most important building blocks of AI and their use cases in the context of financial services.

1. Machine Learning

Machine learning refers to the ability of software to learn on its own without being programmed. Machine learning programs adjust their algorithms in response to new insights. Where data mining algorithms would hand over findings to human beings for further work, machine learning can act on its own.

Banks and credit unions can use machine learning across the front, middle and back office, in functions ranging from customer service to sales and marketing to fraud detection to securities settlement. For instance, in the middle office, they can identify or even prevent fraud by deploying machine learning to look into patterns in payment transaction data to spot anomalies or inconsistencies.

2. Deep Learning

Deep learning leverages a hierarchy of artificial neural networks, similar to those in the human brain, to do its job. Unlike traditional programs, which think linearly, deep learning mimics the human brain to perform non-linear deductions. Deep learning systems produce better decisions by factoring learning from previous transactions or interactions to draw conclusions. For example, they can gather information about customers and their behaviors from social networks and from that infer their likes and preferences. Financial institutions can use this insight to make contextual, relevant offers to those customers in real-time.

3. Natural Language Processing

Natural Language Processing (NLP) is a key building block that will help computers learn, analyze, and understand human language. NLP can be put to use to organize and structure knowledge in order to answer queries, translate content from one language to another, recognize individual people by their speech, mine text, and perform sentiment analysis. Besides improving customer service, natural language processing-capable systems will, over time, learn to resolve issues automatically.

Banking providers have started to leverage NLP in different ways. On the website of the largest bank in the United States, virtual assistants offer support for credit cards, loans and other banking services. Singapore’s DBS Bank uses a virtual assistant called KAI to enhance the experience at Digibank, its mobile-only bank in India. KAI helps Digibank to anticipate and answer thousands of customer queries, and assist customers with their banking transactions in real time.

4. Natural Language Generation

Natural Language Generation (NLG) is also a foundational AI technology. Wherein NLP will help computers analyze, understand, and make sense of human language, NLG will help them to converse and interact intelligently with humans. Banks mainly leverage NLG for purposes that require data from multiple sources to be combined to produce insights in a format that is easily understood. NLG can also knit raw data into a narrative, which banks such as Credit Suisse are using to generate portfolio reviews.

5. Visual Recognition

Visual recognition is a branch of AI that recognizes images and their content. It uses deep learning to perform its role of finding faces, tagging images, identifying the components of visuals, and picking out similar images from a large set. Visual recognition feeds off huge amounts of data and needs open source software libraries and frameworks to function well. A key application of visual recognition technology in banking will be similar to that of speech recognition — enabling a frictionless customer experience.

To this end, several banks have adopted visual recognition in common front-end operations. Australia’s Westpac, for example, is using the technology to allow customers to activate a new card from their smartphone camera, while Santander is one of those using it to authenticate documents. And many other banks, including Bank of America, Citibank, Wells Fargo, and TD Bank, are leveraging this technology to allow customers to deposit checks remotely via mobile.

The post The Five Fundamental Building Blocks for Artificial Intelligence in Banking appeared first on Artificial Intelligence.

Last week, the credit reporting agency Equifax announced that malicious hackers had leaked the personal information of 143 million people in their system. That’s reason for concern, of course, but if a hacker wants to access your online data by simply guessing your password, you’re probably toast in less than an hour. Now, there’s more bad news: Scientists have harnessed the power of artificial intelligence (AI) to create a program that, combined with existing tools, figured more than a quarter of the passwords from a set of more than 43 million LinkedIn profiles. Yet the researchers say the technology may also be used to beat baddies at their own game.

The work could help average users and companies measure the strength of passwords, says Thomas Ristenpart, a computer scientist who studies computer security at Cornell Tech in New York City but was not involved with the study. “The new technique could also potentially be used to generate decoy passwords to help detect breaches.”

The strongest password guessing programs, John the Ripper and hashCat, use several techniques. One is simple brute force, in which they randomly try lots of combinations of characters until they get the right one. But other approaches involve extrapolating from previously leaked passwords and probability methods to guess each character in a password based on what came before. On some sites, these programs have guessed more than 90% of passwords. But they’ve required many years of manual coding to build up their plans of attack.

The new study aimed to speed this up by applying deep learning, a brain-inspired approach at the cutting edge of AI. Researchers at Stevens Institute of Technology in Hoboken, New Jersey, started with a so-called generative adversarial network, or GAN, which comprises two artificial neural networks. A “generator” attempts to produce artificial outputs (like images) that resemble real examples (actual photos), while a “discriminator” tries to detect real from fake. They help refine each other until the generator becomes a skilled counterfeiter.

Giuseppe Ateniese, a computer scientist at Stevens and paper co-author, compares the generator and discriminator to a police sketch artist and eye witness, respectively; the sketch artist is trying to produce something that can pass as an accurate portrait of the criminal. GANs have been used to make realistic images, but have not been applied much to text.

The Stevens team created a GAN it called PassGAN and compared it with two versions of hashCat and one version of John the Ripper. The scientists fed each tool tens of millions of leaked passwords from a gaming site called RockYou, and asked them to generate hundreds of millions of new passwords on their own. Then they counted how many of these new passwords matched a set of leaked passwords from LinkedIn, as a measure of how successful they’d be at cracking them.

On its own, PassGAN generated 12% of the passwords in the LinkedIn set, whereas its three competitors generated between 6% and 23%. But the best performance came from combining PassGAN and hashCat. Together, they were able to crack 27% of passwords in the LinkedIn set, the researchers reported this month in a draft paper posted on arXiv. Even failed passwords from PassGAN seemed pretty realistic: saddracula, santazone, coolarse18.

Using GANs to help guess passwords is “novel,” says Martin Arjovsky, a computer scientist who studies the technology at New York University in New York City. The paper “confirms that there are clear, important problems where applying simple machine learning solutions can bring a crucial advantage,” he says.

Still, Ristenpart says “It’s unclear to me if one needs the heavy machinery of GANs to achieve such gains.” Perhaps even simpler machine learning techniques could have assisted hashCat just as much, he says. (Arjovsky concurs.) Indeed, an efficient neural net produced by Carnegie Mellon University in Pittsburgh, Pennsylavania, recently showed promise, and Ateniese plans to compare it directly with PassGAN before submitting his paper for peer review.

Ateniese says that though in this pilot demonstration PassGAN gave hashCat an assist, he’s “certain” that future iterations could surpass hashCat. That’s in part because hashCat uses fixed rules and was unable to produce more than 650 million passwords on its own. PassGan, which invents its own rules, can create passwords indefinitely. “It’s generating millions of passwords as we speak,” he says. Ateniese also says PassGAN will improve with more layers in the neural networks and training on many more leaked passwords.

He compares PassGAN to AlphaGo, the Google DeepMind program that recently beat a human champion at the board game Go using deep learning algorithms. “AlphaGo was devising new strategies that experts had never seen before,” Ateniese says. “So I personally believe that if you give enough data to PassGAN, it will be able to come up with rules that humans cannot think about.”

The post Artificial intelligence just made guessing your password a whole lot easier appeared first on Artificial Intelligence.

MICROSOFT SAID LAST week that it had achieved a new record for the accuracy of software that transcribes speech. Its system missed just one in 20 words on a standard collection of phone call recordings—matching humans given the same challenge.

The result is the latest in a string of recent findings that some view as proof that advances in artificial intelligence are accelerating, threatening to upend the economy. Some software has proved itself better than people at recognizing objects such as cars or cats in images, and Google’s AlphaGo software has overpowered multiple Go champions—a feat that until recently was considered a decade or more away. Companies are eager to build on this progress; mentions of AI on corporate earnings calls have grown more or less exponentially.

Now some AI observers are trying to develop a more exact picture of how, and how fast, the technology is advancing. By measuring progress—or the lack of it—in different areas, they hope to pierce the fog of hype about AI. The projects aim to give researchers and policymakers a more clear-eyed view of what parts of the field are advancing most quickly and what responses that may require.

“This is something that needs to be done in part because there’s so much craziness out there about where AI is going,” says Ray Perrault, a researcher at nonprofit lab SRI International. He’s one of the leaders of a project called the AI Index, which aims to release a detailed snapshot of the state and rate of progress in the field by the end of the year. The project is backed by the One Hundred Year Study on Artificial Intelligence, established at Stanford in 2015 to examine the effects of AI on society.

Claims of AI advances are everywhere these days, coming even from the marketers of fast food and toothbrushes. Even boasts from solid research teams can be difficult to assess. Microsoft first announced it had matched humans at speech recognition last October. But researchers at IBM and crowdsourcing company Appen subsequently showed humans were more accurate than Microsoft had claimed. The software giant had to cut its error rate a further 12 percent to make its latest claim of human parity.

The Electronic Frontier Foundation, which campaigns to protect civil liberties from digital threats, has started its own effort to measure and contextualize progress in AI. The nonprofit is combing research papers like Microsoft’s to assemble an open source, online repository of data points on AI progress and performance. “We want to know what urgent and longer term policy implications there are of the real version of AI, as opposed to the speculative version that people get overexcited about,” says Peter Eckersley, EFF’s chief computer scientist.

Both projects lean heavily on published research about machine learning and AI. For example, EFF’s repository includes charts showing rapid progress in image recognitionsince 2012—and the gulf between machine and humanperformance on a test that challenges software to understand children’s books. The AI Index project is looking to chart trends in the subfields of AI getting the most attention from researchers.

The AI Index will also try to monitor and measure how AI is being put to work in the real world. Perrault says it could be useful to chart the numbers of engineers working with the technology and the investment dollars flowing to AI-centric companies, for example. The goal is to “find out how much this research is having an impact on commercial products,” he says—although he concedes that companies may not be willing to release the data. The AI Index project is also working on tracking the volume and sentiment of media and public attention to AI.

Perrault says the project should win a broad audience because researchers and funding agencies will be keen to see which areas of AI have the most momentum, or need for support and new ideas. He says banks and consulting companies have already called, seeking a better handle on what’s real in AI. The tech industry’s decades-long love affair with Moore’s Law, which measured and forecast advances in computer processors, suggests charts showing AI progress will find a ready audience in Silicon Valley.

It’s less clear how such measures might help government officials and regulators grappling with the effects of smarter software in areas like privacy. “I’m not sure how useful it’ll be,” says Ryan Calo, a law professor at the University of Washington who recently proposed a detailed roadmap of AI policy issues. He argues that decisionmakers need a high-level grasp of the underlying technology, and a strong sense of values, more than granular measures of progress.

Eckersley of the EFF argues that AI tracking projects will become more useful with time. For example, debate about job losses might be informed by data on how quickly software programs are advancing to automate the core tasks of certain workers. And Eckersley says looking at measures of progress in the field has already helped convince him of the importance of work on how to make AI systems more trustworthy. “The data we’ve collected supports the notion that the safety and security of AI systems is a relevant and perhaps even urgent field of research,” he says.

Researchers in academia and at companies such as Google have recently investigated how to trick or booby-trap AI software and prevent it from misbehaving. As companies rush to put software in control of more common technology such as cars, measurable progress on how to make it reliable and safe could be the most important of all.

The post DO WE NEED A SPEEDOMETER FOR ARTIFICIAL INTELLIGENCE? appeared first on Artificial Intelligence.

Whether it’s reports of a new and wondrous technological accomplishment or of the danger we face in a future filled with unbridled machines, artificial intelligence (AI) has recently been receiving a great deal of attention. If you want to understand what the fuss is all about, Max Tegmark’s original, accessible, and provocative Life 3.0: Being Human in the Age of Artificial Intelligence would be a great place to start.

The book’s goal is not to tell us what being human will look like in the years ahead, as the title might seem to suggest, but rather to give us the background necessary to understand where technology might lead the human species. In this it succeeds, bringing well-timed clarity to the sometimes muddled public view of AI that has emerged over the past few years.

When computer scientist John McCarthy gave the field its name in 1955, AI’s scholars grappled with the tantalizing prospect that computers might have the capacity to demonstrate broad human-level intelligence, something that is now increasingly called “artificial general intelligence” (AGI). Achieving AGI, however, proved difficult, and researchers were forced to strategically target more narrow tasks, focusing on problems such as understanding images, interacting with natural language, manipulating objects in the physical world, learning, and even playing games. The timeliness of Life 3.0arises from the unprecedented number and range of successes seen in these areas in just the past few years and the ensuing publicity these successes have generated.

Recent depictions of the future of AI run the gamut from benevolent machines letting people live lives of leisure to nightmarish slaughterers of the human race. Part of this dichotomy may come from the fact that not everyone means the same thing when they refer to AI. Some are focused on today’s progress and the potential implications of AI automation (see, for example, Erik Brynjolfsson and Andrew McAfee’s The Second Machine Age). Others, however, are talking about AGI (as in Nick Bostrom’s Superintelligence).

Tegmark successfully gives clarity to the many faces of AI, creating a highly readable book that complements The Second Machine Age’s economic perspective on the near-term implications of recent accomplishments in AI and the more detailed analysis of how we might get from where we are today to AGI and even the superhuman AI in Superintelligence.

Tegmark begins by laying out the range of perspectives currently found among those working in the field. He showcases, especially, the increasingly mainstream view that we should be thinking more deeply about the societal implications of what we create and how we might ultimately design and build AI systems that reflect and respect our hopes and values.

Life 3.0 focuses both on the short-term status of AI and on AGI and the longer-term outlook, projecting from tens or hundreds of years to tens of thousands of years to billions of years ahead. Along the way, Tegmark gives us a physicist’s take on intelligence and computation, the origin and nature of goal-oriented behavior and its implications for AGI, the nature of consciousness and what it might mean for AGI, and—veering from the book’s main focus on AI—what limits physics might impose on our future.

Life 3.0 is interlaced with these and many other thought-provoking ideas. For example, are feelings a consequence of the universe maximizing its entropy? Just as computers work on a “substrate” of 0s and 1s independent of their precise physical implementation in the machine, is intelligence similarly “substrate-independent,” with the potential for implementation not just in biological neurons but also computer hardware? Would intelligent machines that design and build new iterations of themselves represent a new form of “life” (the “Life 3.0” Tegmark refers to in the book’s title)? The book is also populated with numerous science fiction–worthy futures—both good and bad—built on a range of optimistic and often provocative projections of where technology may go and may lead us.

At one point, Tegmark quotes Emerson: “Life is a journey, not a destination.” The same may be said of the book itself. Enjoy the ride, and you will come out the other end with a greater appreciation of where people might take technology and themselves in the years ahead.

The post A physicist explores the future of artificial intelligence appeared first on Artificial Intelligence.