CES 2018: When will AI deliver for humans?

In Las Vegas this week you can learn a lot about the exciting potential of artificial intelligence. You can also be left wondering whether AI is a triumph of marketing, yet to deliver real improvements to the economy and the way we live.

One of my first stops here was at a University of Las Vegas robotics lab. Scientists there were working on projects ranging from drones to virtual reality, but they were also collaborating with the team behind one of the stars of the robot world,

Sophia is a humanoid robot with a face that strays dangerously close to that "Uncanny Valley" where machines look just a little too human for comfort. She achieved fame and some notoriety when Saudi Arabia named her its first robot citizen, a stunt which saw many question whether machines were now being given freedoms denied to Saudi women.

Sophia is the work of Hanson Robotics, an American firm with a base in Hong Kong. We had come to see her take her first halting steps on legs which are the work of this university lab.

Then it was time for a handshake and a chat, with Sophia responding remarkably articulately to my questions - though I should say she'd had advance notice of what I'd ask.

In many ways, this a hugely impressive project combining expertise in robotics, speech recognition and machine learning developed by American and Chinese scientists.

But when I ask what practical purpose she serves, Hanson Robotics founder David Hanson is somewhat vague, mentioning work with autistic children and a role as a work of art.

He is, however, a man with a startling ambition to use Sophia as a platform to achieve the holy grail, Artificial General Intelligence where machines can outperform humans at any intellectual task.

He admits she is, in many ways, about as intelligent as an amoeba at present, but is hopeful that as she interacts with humans she will learn and grow. "Our aspiration is to bring the machines to life," he tells me.

It's a vision that will be met with scepticism by scientists who think the humanoid path is the wrong route for AI, and with fear from those nervous about killer robots.

Elsewhere we find more immediate attempts to turn AI into saleable products. Voice-controlled devices are everywhere, with Amazon's Alexa now facing a more sustained challenge from the Google and its Assistant. The two giants seem convinced that our homes will soon echo to the sound of people yelling orders at everything from the light switches to the microwave.

And out on the Las Vegas streets the battle to prove self-driving cars are imminent is in full swing. I hitched a ride in an autonomous taxi, a collaboration between Uber rival Lyft and the technology firm Aptiv which has bought up AI teams from both Carnegie Mellon and MIT.

The autonomous driving mode was impressive, keeping a close eye on pedestrians, but as soon as we arrived at Caesars Palace the human driver had to take over - the casinos are still cautious about allowing this kind of technology on their land.

In downtown Las Vegas, I then hopped into a more futuristic vehicle, a taxi with no driver, no steering wheel or pedals. It made its way around a block of streets set aside for autonomous vehicles, but it was not the smoothest ride, with the vehicle lurching to a halt at any hint of an obstacle. Still, Navya the French transport company behind this project, insists its Autonom taxi will be in service in cities in the United States and Australia very soon,

But the most confident statement about artificial intelligence at the conference was made by Baidu. Its AI supremo Qi Lu took to a stage to explain to a Las Vegas audience what his company did - "China's Google" - and why his country was going to close the gap with its American rivals. It was all about sheer scale - China had far more people, producing far more data and there was what he described as a "friendly policy environment".

That presumably means controversial technologies such as facial recognition can be deployed more quickly, creating a virtuous circle where ever more data makes the AI systems improve ever faster.

But as this data gold rush continues and the hype around AI gets ever louder, it is worth remembering the great paradox - in a time of great technological change productivity growth has ground to a halt.

Ever waited ages at an airport to "drop" your bag after thinking you'd done the work by checking in online? Or stood at a hotel reception desk as endless pieces of data are entered into a computer and thought wistfully of the days when you signed a form and were handed a key?

Then you have experienced that productivity paradox and may be sceptical that smart machines learning from vast amounts of data will make our lives more simple and productive.