The China Mail - Rise of the robots: the promise of physical AI

USD -
AED 3.67315
AFN 63.503991
ALL 83.375041
AMD 377.180403
ANG 1.790083
AOA 917.000367
ARS 1383.990604
AUD 1.452433
AWG 1.8
AZN 1.70397
BAM 1.69972
BBD 2.014322
BDT 122.712716
BGN 1.709309
BHD 0.377349
BIF 2968.5
BMD 1
BND 1.28787
BOB 6.936019
BRL 5.255304
BSD 1.000117
BTN 94.794201
BWP 13.787919
BYN 2.976987
BYR 19600
BZD 2.011341
CAD 1.38995
CDF 2282.50392
CHF 0.798523
CLF 0.023433
CLP 925.260396
CNY 6.91185
CNH 6.92017
COP 3680.29
CRC 464.427092
CUC 1
CUP 26.5
CVE 96.12504
CZK 21.309304
DJF 177.720393
DKK 6.492704
DOP 59.72504
DZD 133.275765
EGP 52.642155
ERN 15
ETB 156.62504
EUR 0.866104
FJD 2.260391
FKP 0.75231
GBP 0.75375
GEL 2.680391
GGP 0.75231
GHS 10.97039
GIP 0.75231
GMD 73.503851
GNF 8780.000355
GTQ 7.653901
GYD 209.354875
HKD 7.82605
HNL 26.510388
HRK 6.545204
HTG 131.099243
HUF 338.020388
IDR 16990.8
ILS 3.13762
IMP 0.75231
INR 94.864204
IQD 1310
IRR 1313250.000352
ISK 124.760386
JEP 0.75231
JMD 157.422697
JOD 0.70904
JPY 160.29904
KES 129.903801
KGS 87.450384
KHR 4012.00035
KMF 428.00035
KPW 899.886996
KRW 1508.00035
KWD 0.30791
KYD 0.833446
KZT 483.490125
LAK 21900.000349
LBP 89550.000349
LKR 315.037957
LRD 183.625039
LSL 17.160381
LTL 2.95274
LVL 0.60489
LYD 6.375039
MAD 9.344504
MDL 17.566669
MGA 4175.000347
MKD 53.384435
MMK 2102.490525
MNT 3571.507434
MOP 8.069509
MRU 40.120379
MUR 46.770378
MVR 15.450378
MWK 1737.000345
MXN 18.121104
MYR 3.924039
MZN 63.950377
NAD 17.160377
NGN 1383.460377
NIO 36.720377
NOK 9.70286
NPR 151.667079
NZD 1.740645
OMR 0.385081
PAB 1.000109
PEN 3.459504
PGK 4.309039
PHP 60.550375
PKR 279.203701
PLN 3.72275
PYG 6538.855961
QAR 3.65325
RON 4.427304
RSD 101.818038
RUB 81.419514
RWF 1461
SAR 3.752351
SBD 8.042037
SCR 14.429246
SDG 601.000339
SEK 9.47367
SGD 1.292804
SHP 0.750259
SLE 24.550371
SLL 20969.510825
SOS 571.503662
SRD 37.601038
STD 20697.981008
STN 21.35
SVC 8.75063
SYP 111.824334
SZL 17.160369
THB 32.860369
TJS 9.556069
TMT 3.5
TND 2.926038
TOP 2.40776
TRY 44.433404
TTD 6.795201
TWD 32.044404
TZS 2576.487038
UAH 43.837189
UGX 3725.687866
UYU 40.481115
UZS 12205.000334
VES 467.928355
VND 26337.5
VUV 119.756335
WST 2.77551
XAF 570.070221
XAG 0.014291
XAU 0.000222
XCD 2.70255
XCG 1.802452
XDR 0.706792
XOF 568.000332
XPF 104.103591
YER 238.603589
ZAR 17.119995
ZMK 9001.203584
ZMW 18.826586
ZWL 321.999592
  • RBGPF

    -13.5000

    69

    -19.57%

  • CMSD

    -0.0900

    22.66

    -0.4%

  • GSK

    -0.1000

    53.84

    -0.19%

  • RIO

    0.8500

    86.64

    +0.98%

  • VOD

    -0.1400

    14.49

    -0.97%

  • NGG

    -0.4800

    81.92

    -0.59%

  • BCE

    -0.2200

    25.25

    -0.87%

  • CMSC

    -0.0500

    22.77

    -0.22%

  • RYCEF

    -0.5900

    14.65

    -4.03%

  • RELX

    -0.1000

    31.97

    -0.31%

  • AZN

    5.0200

    188.42

    +2.66%

  • JRI

    -0.2700

    11.8

    -2.29%

  • BCC

    0.1400

    74.43

    +0.19%

  • BTI

    0.3749

    57.8

    +0.65%

  • BP

    0.5100

    46.68

    +1.09%

Rise of the robots: the promise of physical AI
Rise of the robots: the promise of physical AI / Photo: © AFP

Rise of the robots: the promise of physical AI

A pair of swivelling, human-like robotic arms, built for physical artificial intelligence research, mirror the motions of an operator in a VR headset twirling his hands like a magician.

Text size:

With enough practice, arms like these can complete everyday tasks alone, says Tokyo company Enactic, which is developing humanoid robots to wash dishes and do laundry in short-staffed Japanese care homes.

Welcome to the future of AI as it starts to infiltrate the material world in the form of smart robots, self-driving cars and other autonomous machines.

"The next wave of AI is physical AI," Jensen Huang, head of US chip giant Nvidia, said last year.

That's "AI that understands the laws of physics, AI that can work among us" and understands "how to perceive the world", Huang added.

Tech firms are pouring massive sums into physical AI, and Morgan Stanley predicts the world could have more than a billion humanoid robots by 2050.

The buzz is only heightened by videos showing advanced androids, often Chinese-made, dancing to Taylor Swift or pulling heavy objects with ease.

Beyond the promise of sci-fi robot butlers, the race has sparked concern over job losses, privacy and how long these innovations will take to actually be useful.

Hiro Yamamoto is the 24-year-old CEO of Enactic, whose OpenArm physical AI training devices are used by Nvidia and at top universities such as Stanford.

He plans to begin deploying new robots, currently under development, from next summer to "live alongside people in environments that are very chaotic, and where conditions are always changing" like care homes.

"So it has to be safe," with a soft exterior that won't injure anyone, Yamamoto said.

- 'Any human role' -

In the Chinese city of Guangzhou, a female figure with a glowing oval-shaped visor for a face, clad in white woven fabric like a fencing athlete, walked slowly across a stage last week to cheers and whispers.

It was the latest humanoid robot to be unveiled by Chinese electric vehicle maker XPeng, which is also pushing into physical AI.

Nimble machines made by US companies, such as Boston Dynamics' dog-like robots, have grabbed headlines over the years.

But government support and strong domestic supply chains are helping Chinese rivals, also including Unitree Robotics and EngineAI, race ahead.

"I haven't given much thought to how many robots we will sell annually in 10 years' time, but I think it would be more than cars," XPeng CEO He Xiaopeng told reporters.

XPeng's robots walk and even dance autonomously -- but how well they handle objects, a more complicated feat, has not been widely demonstrated.

Their dexterous fingers and flexible skin are unlikely to replace workers on China's factory floors soon, He said.

The cost of one robot hand, which needs to be replaced regularly for heavy-duty work, could pay a Chinese worker's salary for years.

But with enough data and training, AI humanoid robots could one day perform "almost any human role", from nanny to home chef or gardener, XPeng co-president Brian Gu told AFP.

- On-the-job training -

Text-based AI tools like ChatGPT are trained on huge volumes of words, but physical AI models must also grapple with vision and the spatial relationship between objects.

For now, remotely operating AI robots to teach them how to do something like picking up a cup "is by far the most reliable way to collect data", Yamamoto said.

Just 30 to 50 demonstrations of each task are needed to fine-tune "vision-language-action" AI models, he added.

Enactic has approached several dozen care facilities in Japan to propose that its teleoperated robots take over menial tasks, so qualified care workers have more time to look after elderly residents.

This on-the-job experience will train physical AI models so the robots can act autonomously in future, Yamamoto said.

US-Norwegian startup 1X is taking a similar approach for its humanoid home helper NEO, which it will deliver to American homes from next year.

NEO costs $20,000 to buy, but so far its performance is shaky, with one video in US media showing the robot struggling to close a dishwasher door, even when teleoperated.

- Physical limits -

In another embarrassing moment, a Russian humanoid robot, said to be the country's first, staggered then fell flat on its face as it made its debut on stage earlier this week.

There is currently a "big gap" between robots' AI systems and their physical abilities, which lag behind, said Sara Adela Abad Guaman, assistant professor in robotics at University College London.

"Nature has shown us that in order to adapt to the environment, you need to have the right body," Abad told AFP, giving the example of a mountain goat that stumbles on ice.

Nevertheless, big deals are being struck, even as booming investment in artificial intelligence feeds fears of a stock market bubble.

Japan's SoftBank recently called physical AI its "next frontier" as it said it was buying industrial robot maker ABB Robotics for $5.4 billion.

Automation raises questions about the future of human labour, but Abad is not too worried.

At the end of the day, "our sense of touch is incomparable," she said.

T.Wu--ThChM