Latest / In The Loop / AI’s Just Made Robots Interesting Again

Transcript

AI’s Just Made Robots Interesting Again · In The Loop

• 0:00
  For over 60 years, we've been promised robots that can do
• 0:03
  almost anything. I think robotics has given
• 0:05
  itself a reputation of being one of the biggest technology over
• 0:09
  promises ever. But right now, something's
• 0:12
  different. In this month, October 2025 / 6
• 0:15
  billion dollars have been poured into robotics by some of the
• 0:18
  biggest names in investing, and they've claimed that their
• 0:22
  future is physical AI. By the end of today's episode,
• 0:26
  you'll start to understand what's actually changed, why
• 0:29
  people are putting so much money into robotics all of a sudden,
• 0:32
  and whether we're at the beginning of an iPhone era
• 0:35
  moment or just the start of another period of
• 0:39
  disappointment. This is another loop with Jack
• 0:41
  Horton. I hope you enjoy the show.
• 0:55
  So let's start by, I guess, setting the scene a little.
• 1:00
  I think ever since the Jetsons in I think the 1960s or 70s had
• 1:05
  this like robot made, it's been promised by I guess
• 1:10
  technologists and robot makers that these are the future of
• 1:13
  household helpers. And in fact, I think experts,
• 1:16
  you know, ever since the 1950s have been pretty confident in
• 1:20
  predicting that, you know, by the year 2000, most American
• 1:23
  homes will have some form of automation to make coffee or
• 1:27
  close windows. But I don't think we've ever
• 1:29
  reached that cool and exciting future because none of this has
• 1:33
  happened, really, other than the ability, I guess, to use a
• 1:35
  mobile phone app to shut your windows in certain hotels.
• 1:38
  And also, you can't really forget the awful, like, robot
• 1:43
  hoovers and lawn mowers that you get in some places.
• 1:45
  So then why all of a sudden is someone pouring $6 billion into
• 1:50
  robotics ventures? Because that's what's happened
• 1:52
  this month. SoftBank pretty famous for
• 2:07
  throwing away money at times, but they've made the headlines
• 2:10
  recently because they've put $5.4 billion alone to acquire a
• 2:15
  massive robot division with the CEO of SoftBank saying that
• 2:19
  Softbank's next frontier is physical AI.
• 2:22
  And Elon Musk in the last month for XAI, such as AI startup
• 2:26
  that's inside of X and Twitter has actually started hiring top
• 2:31
  NVIDIA researchers to develop what is described as world
• 2:34
  models. So these are like AI language
• 2:37
  models and systems that I guess give robots an intuitive
• 2:41
  understanding of physical world objects and gravity and
• 2:45
  basically how the world just works.
• 2:47
  There's also been big humanoid robots making strides in the
• 2:50
  world. For anyone that's been on
• 2:51
  Instagram and is connected to technology accounts, really
• 2:55
  you've just seen the robot called Figure.
• 2:58
  So Figure O3 is a humanoid robot that's basically aimed at
• 3:02
  households to do general chores. And in their videos, you can go
• 3:06
  YouTube. This Figures robot has been, you
• 3:08
  know, picking up clutter or taking clothes out the washer
• 3:12
  and loading dishwashers. And obviously as a result, a lot
• 3:15
  of the ecosystem around robotics is racing to support these
• 3:18
  smarter robots. So NVIDIA have announced Project
• 3:22
  Groot. So Groups is basically platform
• 3:24
  for training robots, they call it for robot intelligence.
• 3:28
  And CEO of NVIDIA, Jensen Huang actually called this the
• 3:32
  foundation for making robots. One of the most exciting
• 3:35
  problems to solve in AI today. And there are lots of new
• 3:39
  breakthroughs coming out to allow robotics to take giant
• 3:43
  leaps forward because of AI. So as you can imagine, Nvidia's
• 3:46
  teamed up with basically every single robotics firm in the
• 3:49
  world from Agility to Boston Dynamics to start-ups like
• 3:53
  Figure and Sanctuary. So obviously there's a, there's
• 3:56
  a renewed focus on robotics right now.
• 4:00
  And there's a good question like why, why is this happening?
• 4:02
  You know, why is this time different than all the others?
• 4:05
  And I think it's a really interesting question to ask.
• 4:07
  And in one word, it's because of AI.
• 4:09
  Not the term AI that used to be used over the last few decades,
• 4:13
  but the specific recent breakthroughs in large language
• 4:16
  models because they've completely transformed the way
• 4:19
  that vision and language is interpreted by machines.
• 4:24
  So the big bet right now is that these same breakthroughs can
• 4:27
  give robotics the adaptability and I guess St. smarts that
• 4:32
  they've historically lacked. So as soon as a human child has
• 4:36
  been birthed into the world, they can learn things by just
• 4:39
  picking up a toy or observing people through trial and error.
• 4:43
  But yet pre twenty 20s robots just cannot do that.
• 4:46
  They're either explicitly coded step by step or trained with
• 4:50
  very narrow data sets. Which means soon as it
• 4:53
  encounters something new IT just gets confused.
• 4:55
  Whereas new large language models, as most of us have
• 4:58
  experienced, are incredibly adaptable and have a huge amount
• 5:02
  of training data. So they've really turned this
• 5:05
  old world robotics, I guess idea on its head into something that
• 5:10
  now is potentially very plausible and exciting.
• 5:12
  For example, Google's DeepMind robotic centre model wasn't
• 5:16
  trained to do 1 specific task. It was trained to integrate all
• 5:19
  of its vision and language capabilities.
• 5:22
  So this robot is able to interpret commands and map them
• 5:25
  to what actions they should take.
• 5:27
  So in a demo, when the robot was asked to hand me a green
• 5:31
  vegetable from the fridge, a Gemini robot arm was able to
• 5:34
  understand this is a vegetable that's green, this is a fridge
• 5:37
  and hand the person the vegetable.
• 5:39
  And coming back to that Xai and Elon Musk's idea of a world
• 5:42
  model. This is where models suddenly
• 5:45
  can teach robots physics. So a world model is essentially
• 5:48
  an internal simulation where AI can start to predict outcomes of
• 5:52
  actions before taking them because they just intuitively
• 5:55
  understand the world. It just adds a massive layer of
• 5:58
  reasoning instead of just reflex or hard coded instructions.
• 6:02
  Another massive breakthrough has been training data and
• 6:06
  exponential data at scale. Obviously companies are now
• 6:10
  using techniques like reinforcement learning in
• 6:12
  virtual environments. So imagine virtual robots in a
• 6:16
  video game that are all essentially characters trying
• 6:19
  tasks in a simulator so they can learn from failures and share
• 6:23
  knowledge, but without actually suffering a negative
• 6:27
  consequence. A good example of this might be
• 6:29
  learning and teaching them how to stack dishes.
• 6:32
  And if they fail, they haven't broken edition real life, it's
• 6:36
  just a video game. But all of that data that
• 6:38
  they're capturing through that trial and error process is
• 6:42
  maintained in its memory for future interactions.
• 6:45
  And this is really exciting because, you know, consider hand
• 6:47
  eye coordination and just general object recognition tasks
• 6:51
  like distinguishing a open door from an open window used to be
• 6:56
  something to be quite difficult for a robot.
• 6:57
  But modern vision models and all this innovations in training
• 7:01
  data makes this type of recognition something that's
• 7:04
  very plausible. And a good example of this is
• 7:07
  that Figures CEO Brett Adcock actually claimed that their
• 7:11
  internal AI robot learnt how to fold a towel from just 80 hours
• 7:15
  of training data, which is something that was unimaginable
• 7:19
  10 years ago. And we also have things like
• 7:21
  integrations and connectivity being 10 times better than say
• 7:24
  1015 years ago. We have cloud, Internet things
• 7:27
  and just frameworks that didn't exist before.
• 7:30
  A good example is a robot could offload really heavy
• 7:33
  computations to the cloud and it could get over the air updates
• 7:37
  like software updates. So when robots learn something
• 7:41
  in the world, they could then force a software update to all
• 7:44
  of them, then learn something. So yeah, there's been a lot of
• 7:46
  interesting breakthroughs and and partially explains the rise
• 7:49
  in excitement and investment levels into the field.
• 7:52
  However, and it is a big however, it's just never that
• 7:55
  simple. Because I think before we get
• 8:07
  carried away with all these exciting demos, it's important
• 8:11
  to note some of the real big challenges because even today's
• 8:15
  most advanced robots are still struggling with basic tasks in
• 8:18
  very controlled environments. A Time reporter note that during
• 8:23
  figures, that robot figure that I mentioned earlier during their
• 8:26
  towel folding test, that robots were caught frequently dropping
• 8:30
  things, being unable to pick the towers up, and having to have an
• 8:33
  engineer come sort things out. I'm not just like criticizing
• 8:36
  them for the sake of it or trying to find gotcha moments.
• 8:39
  These are just real challenges that they really need to
• 8:41
  overcome if this is something that's going to generate revenue
• 8:44
  and deliver value in a house. I think this really speaks to
• 8:47
  hype cycles because historically there's been lots of AI and
• 8:51
  robotics hype cycles. Hype builds big expectations
• 8:54
  form and then obviously reality triggers a big crash to Earth
• 8:59
  and a collapse in funding. We saw this in the late 1880s
• 9:02
  and then again in the 2000s when consumer robots with those, you
• 9:06
  know, rubbish Hoover robots popped up everywhere and
• 9:09
  everyone thinking that was going to be the next big thing.
• 9:11
  But after 2015, when lots of high profile robotics companies
• 9:16
  were shut down, then investors grew quite rare, wary and and
• 9:20
  just refused to put money into the sector.
• 9:22
  Even some of the really exciting companies that you've probably
• 9:24
  seen all over YouTube, Boston Dynamics, you know, that robot
• 9:27
  that's kind of painted yellow, the dog that runs around, the
• 9:30
  robot that does backflips that you can't push over.
• 9:33
  I mean, their videos have been super viral, but they've still
• 9:36
  not made any money. Google bought it in 2013, but
• 9:41
  finding no near term product market fit, they sold it to
• 9:44
  SoftBank in 2017. And SoftBank then sold it to
• 9:48
  Hyundai in 2020 for about $1.1 billion.
• 9:51
  And the real issue is that Boston Dynamics never found any
• 9:54
  mass market applications. They started selling Spot the
• 9:58
  Robot Dog in 2019, but it cost $75,000 each and really it was
• 10:04
  just doing inspections around building sites.
• 10:06
  Atlas, the humanoid robot was the exact same.
• 10:08
  It was a very expensive prototype with no rear clear
• 10:12
  path to being usable by the mass market.
• 10:16
  It's just too early. And so yeah, I think we can sum
• 10:18
  up robotics challenges into 3 core buckets.
• 10:21
  First is real world performance is just completely unproven.
• 10:24
  You know, robotics demos are notorious for being very
• 10:27
  controlled with mock kitchens, engineers, even controllers.
• 10:32
  And a task like unloading a dishwasher that might take us 2
• 10:36
  to 5 minutes might take a robot 20 minutes or even longer to
• 10:41
  train to do it. Assume it doesn't get stuck or
• 10:44
  break all the glass. Second robot is just very
• 10:46
  expensive. You know Agility Robotics has a
• 10:49
  humanoid robot and it costs about 150 K.
• 10:52
  Tesla's Optimus has tried to target A20K price tag, but
• 10:57
  that's going to be years away and it's current prices.
• 11:01
  The return on investment for this general purpose robot is
• 11:04
  completely unproven. We could be decades away from
• 11:07
  this being a reality. And 3rd, the struggle with real
• 11:10
  edge cases and generalizations. A robot might do a really good
• 11:14
  job of folding some towels, but when that towel is wet or if
• 11:20
  it's got lots of tassels on it, then often a robot will get
• 11:23
  quite confused. So it's got a lot of training,
• 11:26
  data and innovation still to take place before a robot can
• 11:31
  really adapt to real world scenarios.
• 11:45
  So yeah, to conclude, if I ask the question of do I think AI
• 11:48
  powered robots are here to stay and ready to empty the
• 11:50
  dishwasher? No.
• 11:53
  After digging through all the evidence, the answer is
• 11:54
  definitely no. Or maybe I'm being too harsh.
• 11:57
  Maybe it's not quite yet, but what I would say is that this
• 12:00
  moment starts to feel a little bit different because there are
• 12:03
  real big technology breakthroughs which overcome
• 12:06
  some of the major challenges robotics has faced over the last
• 12:09
  50-60 years. There's also huge investment
• 12:12
  going into the sector and many companies that were badly burned
• 12:15
  before are starting to come back to the table because I think
• 12:17
  most people all know, acknowledge, recognise that it
• 12:21
  might take five years, it might take 50.
• 12:23
  But robotics is going to be a major leap and one of the most
• 12:28
  influential piece of technology in human society.
• 12:32
  And really, if you think about how technology is typically
• 12:34
  spread, I expect industrial adoption first, then gradual
• 12:38
  home integration over the next 6 to 15 years.
• 12:41
  And I know that's a big window, but it's so hard to tell.
• 12:45
  History then gives us a lot of clues.
• 12:47
  Smartphones went from about 0 to 50% global adoption in about 10
• 12:51
  years. Personal computers took 15 to 20
• 12:53
  years. Even robot vacuums launched in
• 12:56
  2002 took 20 years to reach 14% of US households by 2023.
• 13:02
  So yeah, robots and the technology surrounding them are
• 13:04
  getting bigger and more well funded, with a lot of clever
• 13:07
  people being attracted to the sector.
• 13:09
  But it's still far off. And in many ways, the real
• 13:11
  questions we should start to ask ourselves throughout this next
• 13:14
  period is, are we as a society ready for them?
• 13:18
  I think that's an even bigger and more important question.
• 13:20
  One for another day. But anyway, that's it for today.
• 13:23
  I hope you enjoyed the discussion and I'll see you next
• 13:26
  week.