Harvard Scientist Warns: Is AI Creating Artificial Life?

Amazing Reality

Ready to have your mind blown by the latest in AI, synthetic life, robot dogs, and record-shattering humanoids? Dive into this episode for a wild ride through future tech—plus NASA’s new telescope mapping the universe at lightning speed! Don’t miss out—subscribe to Amazing Reality and tell us which breakthrough amazed you most in the comments! #technology #AI #robots #space #science  👉 This channel was created in collaboration with https://www.youtube.com/@realidadeimpressionante   • The Inspired Business Technology de RomanSenykMusic  0:01 - Introduction and Main Topics Overview 1:08 - Artificial Intelligence and Society Risks 4:51 - Creation of Synthetic Life: Sputsel 6:52 - Limits of Artificial Life 7:25 - Delivery and Mobility Robots 9:42 - Automated Parking Solutions 11:43 - Humanoid Robot Racing Advances 13:55 - NASA's New Roman Space Telescope 16:24 - Microduck: Accessible AI Robot Training

Transcript

00:01And what if we're reaching that point where not only technology but life itself is starting to be redesigned?

00:08A Harvard professor warns that artificial intelligence could lead us to a kind of quote-unquote industrial creation of human

00:15beings,

00:16with algorithms increasingly taking over the decisions in our society.

00:21Meanwhile, other scientists have just built from scratch, according to them, a synthetic cell capable of growing.

00:28Copy its DNA, divide, and transmit this material to its offspring.

00:34Could this be the first artificial life?

00:37And as soon, we'll see a robot dog with wheels that climbs stairs and delivers packages on its own.

00:43Chinese humanoid robot running 100 meters in just 8.64 seconds, surpassing the human record set by Usain Bolt.

00:50And NASA launching a telescope capable of mapping the universe a thousand times faster than Hubble.

00:57So, let's unpack all that and a few more things in new tech.

01:08Hello, impressed viewers.

01:10I'm Carlos Alves and you're watching Amazing Reality.

01:17Artificial intelligence has been presented as a technology capable of transforming practically every area of our society.

01:25But for historian Gilles Lepore, a professor at Harvard University, there is an even greater risk in this transformation.

01:32Allowing technology companies to take on roles that have traditionally belonged to people and democratic institutions.

01:40Lepore describes this process using a rather provocative expression.

01:44A kind of, quote-unquote, the industrial creation of human beings.

01:48The researcher presented her argument during an appearance on Hard Fork, the New York Times podcast,

01:53while discussing the concept of the artificial state, which is the subject of her new book, The Rise and Fall

02:00of the Artificial State.

02:02The idea doesn't mean that machines are literally creating human beings, okay?

02:07Lepore uses the comparison to describe a society in which people are increasingly classified, segmented, and isolated by digital systems

02:16controlled by large corporations.

02:19According to her, some of the functions that used to depend directly on institutions, communities, and human interactions are now

02:27being transferred to technological platforms.

02:29And this already appears in relatively common situations.

02:32When someone needs assistance, for example, by calling a company, they might first encounter an artificial intelligence chatbot saying,

02:40Hello, how are you, mister?

02:41Carlos, that's how it works right now, isn't it?

02:45Information, news, speeches, political discussions, and even election-related recommendations are increasingly going through digital platforms.

02:54For Lepore, the fundamental question is quite simple.

02:57Who decided that this should be the way people are living nowadays?

03:02The concept of the artificial state describes a situation in which decisions and social functions are no longer mediated by

03:10the direct consent of people, but by machines.

03:21The argument is that parts of it are being gradually built as companies come to control an increasing share of

03:28the infrastructure used for communication and the circulation of information.

03:32And this transformation started before the current generation of artificial intelligence.

03:37Okay?

03:38Radio, television, personal computers, the internet, and social networks have already profoundly changed the way information circulates.

03:46But artificial intelligence systems can speed up this process.

03:51Today, AI models can produce political content, select information, answer questions, and participate in decisions that previously required some kind

04:00of human interaction.

04:01In the past, there was an old pat on the back and a basket of basic groceries.

04:06At the same time, these technologies are able to process huge amounts of data collected about their users, also increasing

04:12the possibilities for monitoring.

04:14Monitoring.

04:14And it's in this context that he, Lepore, uses the comparison with industrial farming.

04:20In an article previously published in the New Yorker, she described their artificial state as a kind of industrialization of

04:27public life, marked by declassification and segmentation of people, as well as by isolation and the weakening of human communities.

04:35That's why her criticism is not simply an opposition to the development of artificial intelligence.

04:40The main point is who controls these systems and to what extent decisions that affect society should or ought to

04:48be transferred to machines.

04:51Scientists at the University of Minnesota have developed a synthetic cell capable of performing some of the fundamental processes associated

04:59with living cells within a single system.

05:01And we've already talked about a previous version in a past video here.

05:05Called Sputzel, it can acquire resources, grow, replicate its DNA, divide and transmit genetic material to a new generation.

05:14The difference is that this structure didn't start from a natural cell.

05:18It was built from the ground up using non-living and individually defined chemical components.

05:23Previous projects had already managed to reproduce specific cellular functions in the laboratory.

05:29But the breakthrough of Sputzel is bringing together different functions within the same system and completing the cell cycle.

05:36The basic structure starts with a liposome, a small sphere filled with water and surrounded by a fat membrane.

05:43Inside it, the researchers placed the molecular machinery needed to perform its functions and a synthetic genome of approximately 90

05:50,000 base pairs.

05:51Instead of having a single chromosome, like many natural cells,

05:55the genetic instructions of Sputnik Sputzl are distributed across different DNA molecules.

06:00To grow, this cell also uses a method quite different from the one found in nature.

06:05Small, nutrient-filled liposomes provide the necessary components for its functioning,

06:10including molecules, enzymes, ribosomes and material to increase its membrane.

06:15Sputzl produces proteins that allow fusion with these small compartments.

06:20As it receives these resources, it grows.

06:22Then, its genome is replicated and the division process begins.

06:27The researchers introduced a genetic modification that increased the production of a growth-related protein.

06:33The modified cells grew faster and produced more offspring.

06:38After five generations, they had out-competed the original population,

06:42and this advantage became even greater when the available nutrients were reduced.

06:46This shows that selection and competition can also happen within these synthetic systems.

06:52But there's a very important distinction, okay?

06:55Sputzl shouldn't be interpreted as the definitive creation of artificial life.

07:00The researchers themselves don't claim to have created life.

07:03The system still depends on external health to function

07:06and doesn't have several capabilities found even in relatively simple natural cells.

07:11So, just to clarify, okay?

07:14It's not a living cell created artificially,

07:16but it's an important step toward discovering how far biology can be built from scratch, quote-unquote, right?

07:25These delivery robots with wheels, basically a box on wheels,

07:28can already move along sidewalks and transport small orders.

07:32But there's a problem when they reach the last few meters of the route.

07:36Stairs, curbs, and uneven terrain can still prevent these machines from reaching the customer's door,

07:41which is the desired destination.

07:45Throughout this problem, the American company Robot.com introduced the R-Dog,

07:50a quadruped robot developed specifically to complete this final part of deliveries.

07:55The R-Dog robot combines articulated legs with wheels installed on its feet.

08:00On flat surfaces, the wheels allow it to move quickly and efficiently.

08:05However, when it encounters obstacles, the legs can bend and lift the body,

08:09allowing it to cross curbs and even go up or down stairs.

08:12The robot's body also functions as a cargo compartment.

08:16The space was designed to carry various items, including food orders and regular packagers.

08:22When it arrives at its destination,

08:24RoboDog doesn't necessarily need to wait for someone to pick up the package.

08:28A rear opening in the compartment opens and allows the package to slide out.

08:33After that, the robot can continue to its next delivery.

08:37The company has not yet explained how the system will determine exactly which package should be released

08:42when several packages are being transported at the same time.

08:46Making deliveries is just one of the functions planned for this robot.

08:50It has two digital screens installed on the sides of its body,

08:53which can display advertisements while the robot is moving through the streets.

08:57The ads are managed through the R-Ads platform,

09:00which allows you to remotely send and control the displayed content

09:02while tracking the interactions generated.

09:05The robot also has a conversational audio system,

09:07capable of responding to people in real time.

09:10To control its operation, the R-Dog uses REM,

09:13the same fleet management platform already used by the company's other robots.

09:16And this already allows the new quadruped to be added to the infrastructure

09:21that already controls different machines from robot.com

09:24without having to develop a completely separate operating system.

09:28The production version should also use Field AI's artificial intelligence models

09:32as part of its autonomy system,

09:34helping the robot operate in dynamic and unpredictable environments.

09:38Commercial deployment is scheduled for 2027.

09:42And Hyundai has started testing a new solution

09:45that could really change the way parking lots are organized.

09:48Instead of the driver looking for a spot and maneuvering the vehicle,

09:52or the car doing it by itself,

09:53small robots go underneath the car,

09:56lift its wheels,

09:57and transport the entire vehicle to the spot chosen by the system.

10:00The technology will be tested in real conditions

10:03at a residential complex in Seiyong, South Korea.

10:06Each unit is only 11 centimeters thick,

10:10allowing the equipment to go directly underneath different vehicles

10:13without requiring major changes to the garage.

10:16After positioning themselves,

10:18the robots lift the wheels and start transporting the car.

10:21The system can move vehicles weighing up to 3,400 kYm

10:24and reaches a maximum speed of approximately 4.3 kilometers per age.

10:29LiDAR sensors are used to help the equipment

10:31navigate autonomously through the parking lot.

10:34For the driver, the process is relatively simple.

10:37You see, instead of driving around the garage

10:39looking for a good spot as close as possible,

10:42he leaves the car in a designated area

10:44and can leave the area.

10:46From that moment on,

10:47the system takes over the entire operation.

10:49The robots go underneath the vehicle,

10:52lift its wheels,

10:52and transport the car to the available spot.

10:55When the owner returns,

10:57he can use a kiosk to request the car.

10:59The robots then remove the car from the spot

11:01and bring it back to the delivery area.

11:04But Hyundai is not developing this technology

11:07just to eliminate maneuvering.

11:09According to them,

11:10one of the main goals is to increase the number of cars

11:13that can be stored in the same garage.

11:16In conventional parking lots,

11:17it's necessary to leave space between vehicles

11:20to allow the doors to open,

11:21as well as corridors wide enough

11:23for drivers to maneuver.

11:25Now, when no one needs to get in

11:28or out of the cars inside the parking spaces,

11:30these distances can be greatly reduced.

11:33During the tests,

11:34the companies will analyze factors

11:35such as the average waiting time,

11:37the number of vehicles moved,

11:39and the utilization rate of these robots

11:41throughout the day.

11:43And the Chinese humanoid robot

11:45completed the 100-meter dash

11:46in just 8.64 seconds,

11:49thus setting a new record

11:50during the 2026

11:51World Humanoid Robot Games

11:53held in Beijing.

11:55The one responsible for the result

11:57was Tian Tian Gong Ultra

11:58developed by X-Humanoid.

12:00The record was set in the 100-meter race

12:02for large robots,

12:03held on the last day of the competition.

12:05The time was almost a second

12:06below the 9.58 seconds

12:08recorded by Usain Bolt in 2009,

12:11which is the current men's world record

12:12among humans.

12:13But that comparison doesn't mean

12:15that Tian Tian Gong Ultra

12:17is simply a better sprinter than Bolt,

12:19Okay, robots and humans

12:21compete under different conditions and rules,

12:23and the results are not directly comparable

12:26the way sports records are compared today.

12:28However, the performance shows

12:30how quickly these locomotion systems

12:31are really advancing, you see.

12:33During the competition itself,

12:35Tian Tian Gong Ultra

12:36improved its own time several times.

12:38In the first trials,

12:40the humanoid completed the 100-meter

12:41in 9.39 seconds.

12:43In the semifinals,

12:45it reduced the mark to 8.86 seconds.

12:47And in the final,

12:48it reached 8.64 seconds.

12:51And the difference becomes even greater

12:53when we compare the result

12:55to the performance of this same platform

12:57in the previous year.

12:58In the 2025,

12:59Humanoid World Robot Games

13:00Tian Gong Ultra

13:02won the 100-meter dash

13:04with a time of 21.50 seconds.

13:07In just about a year,

13:08the time needed to cover the same distance,

13:10drop to less than half.

13:12And speed isn't the only challenge.

13:14Making a Humanoid run

13:16requires continuously controlling balance,

13:18leg position,

13:19the impact of the feet against the ground,

13:21and the movement of the rest of the body.

13:24The higher the speed,

13:25the less time there is

13:26to correct any instability.

13:29The track events in the games

13:30were also carried out autonomously,

13:32you know,

13:33without remote control,

13:34with exceptions only

13:35for some obstacle events.

13:37And the next challenge

13:39won't just be making

13:39the Humanoid Robots run even faster.

13:42It will be to combine that speed

13:44with balance,

13:45control,

13:46and the ability to operate reliably

13:48outside of a prepared track.

13:50Remember the robots crashing there?

13:53That also happened

13:54with Tian Tian Gong Ultra,

13:55you see?

13:56And NASA launched

13:57a new space telescope

13:59developed to observe

14:00huge areas of the universe

14:01at a speed that

14:02none of its previous

14:03major observatories could achieve.

14:05Called the next

14:06Nisgrace Roman Space Telescope,

14:08the observatory was launched

14:10on August 30, 2026

14:11aboard a SpaceX Falcon Heavy rocket

14:14from the Kennedy Space Center

14:15in Florida.

14:16Now, the Roman has begun

14:18a journey of approximately

14:19three months

14:20to its operational region,

14:22located about

14:22one and a half million kilometers

14:24from Earth.

14:25The telescope was designed

14:26to investigate

14:27some of the biggest questions

14:28in modern astronomy,

14:29including dark matter,

14:31dark energy,

14:32galaxy formation,

14:33and planets located

14:34outside the solar system.

14:36One of its main features

14:37is the ability

14:38to observe large regions

14:39of the sky very quickly.

14:40According to them,

14:42the Roman has a 2.4 meter

14:43primary mirror,

14:44which is approximately

14:45the same size

14:46as the one used by Hubble.

14:47The big difference

14:48is in its field of view.

14:49Its wide field instrument

14:51uses a camera

14:51of approximately 300 megapixels

14:54and can observe

14:55an area of the sky

14:56at least 100 times larger

14:58than Hubble

14:58in a single exposure

14:59while maintaining

15:00similar resolution.

15:01By combining this field of view

15:03with its ability

15:03to quickly switch

15:04between different regions

15:05of space,

15:06NASA states that the Roman

15:07will be able to survey the sky

15:09up to a thousand times

15:10faster than Hubble.

15:11And this doesn't mean

15:13that the new telescope

15:14will replace Hubble

15:15or even James Webb,

15:17the three observatories,

15:18have different functions.

15:19James Webb was mainly developed

15:21to observe extremely distant objects

15:23and study them

15:24in great detail.

15:26Roman will work more

15:27as a large-scale mapper

15:29of the universe,

15:29identifying huge amounts

15:31of objects and phenomena

15:32that can later be examined

15:34by other telescopes.

15:36And this capability

15:37will also produce

15:38a huge amount of information.

15:39According to NASA,

15:41Roman is expected

15:41to transmit approximately

15:431.4 terabytes of data

15:45every day,

15:46the highest rate ever planned

15:47for an astrophysics mission

15:48by the agency.

15:49And this is the highest rate

15:51ever planned

15:51for an astrophysics mission

15:53by the agency.

15:54Artificial intelligence,

15:55machine learning,

15:56and even scientists

15:57and citizens

15:58are expected

15:59to help analyze this data

16:01and identify objects

16:02that may eventually

16:03deserve further study.

16:05If everything goes as planned,

16:07NASA expects to release

16:08the first images

16:08from Roman in early 2027.

16:11After that,

16:12the observatory

16:12will begin to do

16:13what it was built for,

16:15photographing huge regions

16:16of the sky

16:17and turning billions

16:18of stars and galaxies

16:19into one of the largest maps

16:21ever produced

16:22of our universe.

16:23And Rugging Face

16:25introduced a small robot

16:26called BIPT

16:27created to make

16:28physical AI training

16:29much more accessible,

16:31called Microduck.

16:32The robot is shaped

16:34like a duck,

16:34costs approximately $400

16:36and was developed

16:38in collaboration

16:39with the French company

16:40Polymath Polyrobotic.

16:42But despite its toy-like appearance,

16:44the main purpose

16:44is to serve

16:45as an open platform

16:46for developers

16:47to experiment

16:48with embodied artificial intelligence

16:49and reinforcement learning.

16:51The Microduck

16:52is approximately 25 centimeters tall,

16:54weighs less than 800 grams

16:56and uses 15 motors

16:57distributed throughout its legs,

16:59head and neck.

17:00The system also includes

17:02two inertial measurement units

17:03for balance

17:04as well as a camera,

17:06microphone,

17:06speaker,

17:07Wi-Fi,

17:08Bluetooth

17:08and a small LiDAR sensor.

17:10Its beak is articulated

17:11and also works

17:12like a kind of claw,

17:14allowing it

17:14to manipulate objects.

17:16In demonstrations,

17:17the little robot

17:18is shown walking,

17:19squatting,

17:19sitting,

17:20following a laser pointer,

17:22picking up objects

17:23and even using

17:24small roller skates,

17:25you see.

17:26But one of the most

17:27impressive features

17:28appears when something

17:29goes wrong.

17:30The Microduck

17:31was designed

17:31to actually fall.

17:33When the machine

17:34is small and light,

17:36developers can use

17:37reinforcement learning

17:37to allow the robot

17:38to repeatedly try

17:39to perform a task,

17:41make mistakes

17:42and adjust its behavior

17:43during training.

17:44And when it loses

17:45its balance,

17:46it can get back up

17:47and continue operating.

17:49This approach

17:50is particularly useful

17:51because reinforcement learning

17:52usually relies

17:53on a huge number

17:54of attempts.

17:55And of course,

17:56training a large

17:57humanoid robot

17:57this way is expensive.

17:59It can pose risks

18:00to the equipment itself

18:01and every time it falls,

18:03you can't get it back up

18:04as easily as this one.

18:05With a platform

18:06weighing less than

18:07one kilogram

18:07and relatively inexpensive,

18:09these mistakes

18:10become much less problematic.

18:11According to them,

18:12the idea is not to compete

18:13with large industrial

18:14humanoid robots.

18:15It's to offer

18:16a machine small enough

18:18to fall

18:18without major consequences

18:20and open enough

18:21so that researchers

18:22and developers

18:23can modify

18:23what it is actually learning.

18:25In this case,

18:27making mistakes

18:27isn't necessarily

18:28a failure of the robot.

18:29It's part of the process

18:31used to teach the machines.

18:32If you like this episode,

18:34you know what to do, right?

18:35Leave a like,

18:36check your subscription,

18:37check because YouTube

18:37might have removed it.

18:38It doesn't hurt to check.

18:39Leave your like too.

18:40It really helps our channel,

18:41you know.

18:42Also, remember that

18:43I really appreciate you

18:44for always making it

18:45to the end, okay?

18:46On your screen,

18:47there are two suggestions,

18:48one here and one here.

18:49Just click

18:50and I'll see you

18:51in the next one.

18:51Thanks for joining me

18:52and see you next time,

18:53alright?

18:53Bye-bye.

18:54I'm waiting for you.

18:55Click there.

Category

Transcript

Recommended