Huawei kicked it off a few months ago with its Mate 10, claiming the phone’s NPU (neural processing unit) would enable smart photography and machine translation. In reality, it just means more efficiency, with faster text and image processing.

Every year at CES, one company shows off a new giant television – and the year’s, unsurprisingly, claims to use AI. Samsung says its new flagship 85-inch Q9S TV uses AI to convert low-resolution videos into 8K resolution.

This is pretty much what every flat-screen TV does; it’s called upscaling, which uses an algorithm to adjust screen resolution based on the picture quality of each scene. At least, it used to be called upscaling. AI will now almost certainly become part of the conversation when buying a TV. Is the picture quality any good? Yes, it uses AI. Case closed.

If a TV can claim to use AI for something as rudimentary as creating a watchable image, then it truly is open season for the “new” technology. We’re now in an era where anything that uses an algorithm can call itself AI.

“Devices may well be smart and/or intelligent but that doesn’t mean they are AI,” says Simon Bryant, associate director of consumer electronics at analyst firm Futuresource.

“The intelligence could be algorithms or logic to improve performance or data processing, and analytics to detect trends and provide information back to the user, but it’s still not AI.”

So what is artificial intelligence? It is an area of computer science that examines if humans can teach a computer to think, and the idea has been around since the 1950s. In recent years it has become the subject of much R&D as computer algorithms and processing power have got to the point where they can emulate humans to some extent.

There isn’t – and probably never will be – an artificial general intelligence, i.e. a computer that could successfully carry out any intellectual task of which a human is capable. But it is possible to develop software that can recognise objects in a photo, or recognise human speech and speak or write a reply.

The latter is called natural language processing (NLP), and it is by far the most dominant form of AI in consumer technology products. At CES this year, it is well and truly making its mark in the smart home via voice-activated products and virtual assistants.

From iDevices’ voice-activated light switch and the Vobot Halo smart wake-up light to Moen’s smart shower and a plethora of new smart speakers, products you can talk to are everywhere. At CES there is even a toilet – Kohler’s Numi – that you can flush using your voice.

Most of these gadgets are compatible with Alexa, Amazon’s voice assistant, which so far appears to be the dominant software. However, Google’s attempt to usurp Alexa with its “Hey Google” advertising across Las Vegas is working to some extent.

You can now talk to Google Assistant via an LG or Sony TV (activated via microphones in the remote control), while Google’s hands-free system Android Auto puts similar functionality in the car.

“Google Assistant appears more advanced in its use of AI,” says Bryant, who believes it shows the ability to adapt and learn. “Not all products that have a digital/voice assistant are AI.”

The ability of a device to “learn”, called machine learning, is a result of it being programmed to recognise patterns to categorise data. In photo recognition, for example, this might be through taking a huge number of photos and developing software and algorithms that are able to identify recurring objects.

CES 2018: with tech promising to end congestion, treat cancer and any number of ills, what can’t it do? Keep the lights on in Vegas

The trouble is, creating a computerised “neural network” that fakes a process in the human brain takes a lot of human endeavour; hundreds of thousands of photos have to be categorised before the computer can learn enough to categorise a fresh photo. It also takes a lot of processing power, so much so that it is difficult to transfer that neural network for later use on, say, a smartphone.

However, despite not being obvious at the CES, machine learning is having a lot of success in the background.

Google’s latest AI tool, called DeepVariant, was recently used to sequence human genomes in significantly more detail than humans could manage by presenting the data as an image. It initiated discoveries of mutations that would otherwise have remained hidden.

Similarly at the University of Texas, researchers worked with Google’s AI tools to trawl through data from the Kepler Space Telescope after astronomers had found all they could using their existing methods. They found “extra” planets around distant solar systems that would have otherwise remained undiscovered.

“AI certainly does appear to be at a tipping point this year,” Bryant says, but he points to a rise in the use of data science, analytics, cloud platforms, NLP and speech recognition being done across the board, rather than the products on show at CES.

On the exhibition floor, the direction of AI is heading elsewhere, he says. “I’d give it another year before [AI] becomes so ubiquitous that it becomes tired and gets ditched in favour of a new term. People only switched to AI when ‘smart’ and ‘machine learning’ became overused.”

Facebook finds a way into China by partnering with Xiaomi to introduce virtual reality headsets

What’s the betting that CES 2019 will see marketeers switch to terms like “true AI” and “AI+” to turn our heads?