Is this AI? See if you can spot the technology in your everyday life. – The Washington Post

Artificial intelligence is suddenly everywhere. Fueled by huge technological advances in recent years and gobs of venture capitalist money, AI has become one of the hottest corporate buzzwords.

Roughly 1 in 7 public companies mentioned artificial intelligence in their annual filings last year, according to a Washington Post analysis. But the term is fuzzy.

AI is purposefully ill-defined from a marketing perspective, said Alex Hanna, director of research at Distributed AI Research Institute. It has been composed of wishful thinking and hype from the beginning.

So what is AI, really? To cut through the hype, we asked 16 experts to judge 10 everyday technologies. Try to spot the AI for yourself and see how you compare to readers and the experts.

Chatbots like ChatGPT

Well start with an easy one. The viral chatbot responds to a users prompt by systematically churning out words, producing a surprisingly coherent but not always accurate reply.

Most experts said that the tool is definitely AI, Things are about to get even murkier.

Auto-correct on mobile phones

Most experts said auto-correct could be considered artificial intelligence, and many said it definitely is AI.

Earlier versions of auto-correct guessed what you meant to type by comparing the locations of the keys you pressed with a dictionary of popular words. But the latest versions also consider the context surrounding a word to better predict what you meant, using the same technique at the core of chatbots. Hint: Youll see this technology again.

Story continues below advertisement

Story continues below advertisement

Tap-to-pay credit cards

Not quite. No expert considered tap-to-pay credit cards to be AI. The system uses radio waves to transmit payment information. It sure seems fancy, but its not AI.

Google Translate

Eleven of 16 experts surveyed said Google Translate is definitely AI.

Modern translation services dont simply look up words in a dictionary and return the foreign language match. Rather, the system uses terabytes of multilingual data to build a model of connections between words and their context.

Sound familiar? This technique, called a transformer model, also underlies chatbots and recent auto-correct systems. It provides a more natural, though not always accurate, translation.

Personalized ads

A majority of experts felt it could be AI.

The defining feature of artificial intelligence is that behavior is learned from data rather than explicitly programmed, said Matthew Carrigan, a machine-learning engineer at Hugging Face.

For a non-AI approach, a programmer might write specific rules, like, If a user is a 30-year-old male, show them sports ads. But AI uses machine learning to create its own rules by analyzing large amounts of data. This enables incredibly specific connections, but it can be difficult to understand how a program drew its conclusions.

Story continues below advertisement

Story continues below advertisement

Computer opponents in video games

Youre not alone. Few experts today definitely consider video game opponents to be AI.

Chess has long been a proving ground for AI research. AI researchers often test programs through games because they have well-defined rules with clear victors and losers.

In 1980, a Carnegie Mellon professor offered a $100,000 prize for the first computer program to defeat a world chess champion. Seventeen years later, IBMs Deep Blue beat Garry Kasparov, cementing its place in the history of artificial intelligence.

Beating a world chess champion is no longer impressive. Todays chess engines dominate all human players.

GPS directions

Turn-by-turn navigation uses a defined set of instructions to search though a database of road networks to quickly find the best route.

How much that route-finding technique just carries out preset rules left experts divided on whether GPS could be considered AI. You said

Facial recognition software, like Apple Face ID

The majority of experts consider facial recognition software to definitely be AI, but you said

The technology maps the precise geometry of peoples facial features. Facial recognition has prompted concern from privacy experts as its use is adopted by governments, especially because the technology is less accurate at identifying people with darker skin.

Apples Face ID feature, which lets users unlock their devices by looking at them, projects infrared dots on a face and uses a neural network a mathematical system that acts like a computerized brain to determine whether the face matches.

Microsoft's Clippy

Youre in the minority on this one. Most experts said Clippy is not AI, although 1 in 4 said it could at least be considered AI.

The animated paper clip assistant, launched with Microsoft Office 97, offered tips based on what the user was doing, like offering to help format a letter after Dear Kevin, was typed. But the widely panned feature was a far cry from intelligence. Microsoft turned it off by default in 2001 and launched a self-aware ad campaign that Clippy was out of work.

Story continues below advertisement

Story continues below advertisement

Virtual voice assistants, like Alexa or Siri

Just over half of experts consider a virtual voice assistant to definitely be AI.

It can process words as theyre said and handle requests using machine learning and neural networks, which many experts say have become synonymous with artificial intelligence.

Even among experts, what counts as artificial intelligence is fuzzy.

The term AI has become so broadly used in practice that its almost always better to use a more specific term, said Nicholas Vincent, an assistant professor at Simon Fraser University.

Nothing was unanimously deemed AI by experts, and few products were definitely declared not AI. Most landed somewhere in the middle.

Some experts dont think anything we use today is AI. Current technology is capable of specific tasks they are trained for but dysfunctional at unforeseen events, said Pruthuvi Maheshakya Wijewardena, a data and applied scientist at Microsoft, who identified no product as definitely AI.

The capabilities of an AI is a spectrum, and we are still at the lower end, said Maheshakya Wijewardena.

For Emily M. Bender, a professor of linguistics at the University of Washington, calling anything AI is a way to dodge accountability for its creators.

What artificial intelligence generates, whether its auto-correct, chatbots or photos, is trained from large amounts of data, often pulled off the internet. When that data is flawed, inaccurate or offensive, the results can reflect and even amplify those flaws.

The term AI makes the machines sound like autonomous thinking entities rather than tools that are created and used by people and companies, said Bender.

Emma Kumer contributed to this story.

The experts surveyed were Emily M. Bender, professor, University of Washington; Matthew Carrigan, machine-learning engineer, Hugging Face; Yali Du, lecturer, Kings College London; Hany Farid, professor, UC Berkeley; Florent Gbelidji, machine-learning engineer, Hugging Face; Alex Hanna, director of research, Distributed AI Research Institute; Nathan Lambert, research scientist, Allen Institute for AI; Pablo Montalvo, machine-learning engineer, Hugging Face; Alvaro Moran, machine-learning engineer, Hugging Face; Chinasa T. Okolo, fellow, Center for Technology Innovation at the Brookings Institution; Giada Pistilli, principal ethicist, Hugging Face; Daniela Rus, director, MIT Computer Science & Artificial Intelligence Laboratory; Mahesh Sathiamoorthy, formerly of Google DeepMind; Luca Soldaini, senior applied research scientist, Allen Institute for AI; Nicholas Vincent, assistant professor, Simon Fraser University; and Pruthuvi Maheshakya Wijewardena, data and applied scientist, Microsoft.

Read this article:

Is this AI? See if you can spot the technology in your everyday life. - The Washington Post