Dragonfly’s Titan Landing: NASA’s Quest to Unravel the Mysteries of the Saturnian Moon

[00:00:00] [SPEAKER_00] This is Space Time, Series 29, Episode 117, full broadcast on the 30th of September, 2026. Coming up on Space Time, NASA names Dragonfly's landing site on the Saturnian moon Titan, new clues about the Earth's moon's ancient magnetic field, and why galaxies in the early universe look so weird. All that and more coming up on Space Time.

[00:00:26] [SPEAKER_03] Welcome to Space Time with Stuart Gary.

[00:00:45] [SPEAKER_00] NASA has identified the landing site where its Dragonfly rotocopter will first touch down on the surface of the Saturnian moon Titan. The target area comprises a large dune field with some interdune terrain south of Silic Crater and stretches along the edge of a range of mountains or hills. The International Astronomical Union, the global body responsible for officially designating objects in space, has approved naming the region Amakic Ande. The first part of the name is Mayan.

[00:01:15] [SPEAKER_00] People appeal to the spirit of Amakic to stop strong winds from damaging their crops. The name literally translates into the one who locks up the wind. And Ande aligns with the International Astronomical Union's convention of naming dune fields in Latin after gods and goddesses of wind. Meanwhile, construction work continues on the Dragonfly spacecraft itself, which is now being assembled in a clean room at the Johns Hopkins Applied Physics Laboratory in L'Rell, Maryland.

[00:01:43] [SPEAKER_00] Cables are being carefully installed in the flight fuselage, inviting the Titan-bound rotocraft with its central nervous system. Now collectively, these bundles of wires, cables and connectors make up the spacecraft's electrical harness, which will transmit power and data between the lander's computers, actuators, sensors, scientific instruments and battery. Dragonfly lander harness lead, Jackie Perry, says it's a key milestone for the engineers and technicians building the vehicle.

[00:02:10] [SPEAKER_00] The wire is silver-coated copper insulated with a heat-resistant durable polymer coating and then wrapped in aluminum, which is then attached to plastic and metal connectors. Perry says the wiring harness could only be installed once the flight structure was delivered and the remote interface units and temperature sensors were installed. The harness is typically one of the first components delivered to a spacecraft. Dragonfly passed its critical design review in 2022.

[00:02:37] [SPEAKER_00] Fabrication of the vehicle began in late 2024 and finished last year. Dragonfly is slated for launch in 2028 and will reach the Saturnian moon in late 2034. Dragonfly's operating environment on Titan poses some unique challenges. Because of the rotocraft's thermos bottle design, insulated to retain heat from its nuclear power source so it can stay warm in Titan's extremely cold conditions, the harness had to be designed to route under a layer of thick foam insulation on the outside of the lander

[00:03:06] [SPEAKER_00] and accommodate the circulation of warm air through the inside. Perry says the rotocraft's high power demands require both 4 and 8 gauge wire, yet the harness still had to be flexible enough to weave through the packed interior, which includes the flight systems and instrument boxes, as well as the 136kg battery. Dragonfly will investigate Titan's prebiotic chemistry and assess the moon's habitability.

[00:03:31] [SPEAKER_00] It's not primarily a life detection mission, but aims to study how far organic chemistry has progressed in an environment rich in carbon compounds, which includes possible past mixing of liquid water and organics. The car-sized 875kg lander features twin quadrotor engines powered by a radioisotope thermoelectric generator providing electricity and heat. The 3.85m long vehicle is designed to fly several kilometres between landing sites,

[00:04:00] [SPEAKER_00] allowing it to sample diverse locations. Mission managers expect to cover over 115km during Dragonfly's 3.3 year primary mission, exploring a range of different environments, from organic dunes to deposits associated with an impact crater, silk crater which we mentioned earlier, where liquid water and complex organic materials, which are key to life as we know it, once existed together. Scientific analysis indicates the impact that forms silk crater melted the icy bedrock,

[00:04:28] [SPEAKER_00] potentially creating a large temporary pool of water that could have remained liquid for hundreds of thousands of years under an insulating ice layer like winter ponds on the Earth. This report from NASA TV.

[00:04:41] [SPEAKER_01] NASA's largest moon, Titan, has a thick atmosphere and a frozen surface rich in organic molecules. In 2034, a NASA mission called Dragonfly will arrive at Titan and study its chemical makeup. Dragonfly is a rotorcraft designed to visit multiple sites across the moon's varied terrain. At each new landing site on Titan's surface, Dragonfly uses a pulsed neutron generator and onboard gamma-ray sensor

[00:05:10] [SPEAKER_01] to detect key elements such as carbon and hydrogen in organic materials or oxygen in water ice. Dragonfly determines if there are well-defined layers of these materials just below the lander. For a closer inspection, Dragonfly uses its drill to generate tailings from Titan's hard, frozen surface. Dragonfly is a force of the plant-carousel, chemical and carbon-fluid-like cell. These surface samples can then be ingested through the pneumatic system, carried with Titan air into the chilled sample lines and to the sample collection carousel.

[00:05:40] [SPEAKER_01] One of the carousel sample cups is placed in a pneumatic port. The cup captures the surface material from the cold air stream and transfers it to the chemical laboratory for measurement. Pulses from a laser release large organic molecules from the surface sample for analysis in the mass spectrometer.

[00:05:58] [SPEAKER_01] The mass spectrometer sorts molecules by mass and measures diagnostic fragments that tell dragonfly the kinds of chemical components that are present in the surface, and whether there are molecules of prebiotic interest. For those potential prebiotic samples, a new cup is placed into an oven and heated to release molecules into a gas chromatograph, where they are sorted for size and type before entering the mass spectrometer.

[00:06:23] [SPEAKER_01] This advanced separation of organic components includes isolating molecules with the same formula but different chiral arrangements, or handedness. Having a preference for one handedness over another is a key biosignature for life on Earth. When the chemical analysis is complete, dragonfly may choose to take another surface sample, or find a new location on Titan to investigate.

[00:06:47] [SPEAKER_00] This whole mission is quite a challenge. Titan is unique in our solar system. It has a diameter of 5,150 kilometers, making it Saturn's largest moon, and 50% bigger than the Earth's moon. In fact, it's larger than the planet Mercury. It's also the only moon with a substantial atmosphere, and the only world other than Earth, where clouds release rain that forms streams and rivers, which then flow into lakes and seas.

[00:07:12] [SPEAKER_00] But Titan is so cold, the water there is normally frozen solid, forming bedrock. And instead of water, the liquid rain on Titan is made up of methane and ethane hydrocarbons. But scientists describe Titan as analogous to the early Earth. That's because it preserves a set of environmental and chemical conditions which are thought to have existed on our planet billions of years ago, before life arose, and before free oxygen accumulated in our atmosphere.

[00:07:39] [SPEAKER_00] It therefore serves as a natural laboratory for studying prebiotic chemistry on a planetary scale. Scientists think its complex organic chemistry, methane cycle, and possible environments could support exotic forms of life, or at least prebiotic chemistry. It's also considered one of the more feasible destinations for future human exploration among the outer solar system bodies, due to its thick atmosphere for radiation shielding and available resources.

[00:08:06] [SPEAKER_00] Titan was extensively studied by the Cassini-Huygens mission, which orbited the Saturnian system between 2004 and 2017. The Huygens lander was deployed from the Cassini spacecraft down to Titan's surface on December 25, 2004, landing on January 14, 2005, near the Udiwi region, the boundary between brighter highlands and darker plains. The lander spent about 90 minutes on the surface, taking readings and measurements before the batteries ran out.

[00:08:36] [SPEAKER_00] Huygens was able to confirm that Titan has a thick nitrogen-dominated atmosphere, with methane and a range of aerosols and complex organic chemistry. It measured strong, high-altitude zonal winds on its way down, and weaker winds near the surface. It also measured a surface temperature of minus 179 degrees Celsius, and a surface atmospheric pressure about 50% higher than sea level here on Earth.

[00:09:01] [SPEAKER_00] Intriguingly, it found the ground on Titan was soft and damp, with a consistency of wet sand, clay, or lightly packed snow. Images showed the surface of Titan was littered with water ice pebbles or cobbles a few centimetres across, on an orange-tinted, organic-rich plane. There was also evidence of past liquid flows near the landing site, and possible methane release from the subsurface. Needless to say, as the Dragonfly mission progresses, we'll keep you informed. This is Space Time.

[00:09:32] [SPEAKER_00] Still to come, new clues about the Earth's moon's ancient magnetic field, and we examine why galaxies in the early universe look so weird. All that and more still to come, on Space Time. A new study claims that just like the present-day Earth, our moon once had a geodynamo-driven magnetic field.

[00:10:00] [SPEAKER_00] On Earth, the movement of liquid iron in the planet's outer core generates its global magnetic field. This so-called geodynamo works on a similar principle to a dynamo on a bicycle, which converts mechanical motion into electrical energy. But unlike today's Earth, the moon no longer has a core-generated magnetic field. One of the study's authors, Anna Mittelholz from EDH Zurich, says there's a heated ongoing debate about whether the moon also operated a dynamo in the past.

[00:10:28] [SPEAKER_00] That's because analysis of rock samples brought back to Earth by the Apollo astronauts are somewhat contradictory. Some researchers assume there was a strong magnetic field that existed over a long period of time, somewhere around 4.25 to 3.5 billion years ago, while others find absolutely no evidence of this. You see, in addition to the dynamo theory, there's a second possible explanation for the magnetized lunar rocks, namely impacts from massive meteorites or asteroids, which could have triggered a magnetization process on the moon.

[00:10:57] [SPEAKER_00] But the new research reported in the journal Science Advances now supports the dynamo theory. Mittelholz and colleagues say that around 4.2 billion years ago, the moon did indeed possess an internally generated magnetic field. Now, this claim isn't based on rock samples, but on data collected by probes in lunar orbit, including gravity measurements by NASA's GRAIL probes and magnetic field models drawing on orbital measurements from the Lunar Prospector and CAG-U emissions. To reach their conclusions,

[00:11:27] [SPEAKER_00] the authors examined the D-Y region on the lunar far side, which is never seen from Earth. Mittelholz says D-Y has one of the strongest magnetic field anomalies on the far side of the moon, and a distinct gravity anomaly coincides spatially there as well. That means this region contains rocks that were more strongly magnetized, while at the same time denser than elsewhere. In most cases, the origin of magnetic anomalies measured from lunar orbits are known. But the gravity data gives scientists an insight into the density

[00:11:56] [SPEAKER_00] and consequently the type of material beneath the surface. Mittelholz says, where the magnetic field and gravity signals coincide, it's possible to combine the two and attribute the anomaly to a specific geological feature. And that's where D-Y comes in. It allowed the authors to create an accurate model of the subsurface by jointly processing gravity and magnetic field data. This data shows the area beneath D-Y surface contains a rock body approximately 60 km wide, extending to a depth of around 9 km.

[00:12:27] [SPEAKER_00] It's far denser than the surrounding crust, and it's strongly magnetized. Combined with the surface geochemistry and an arched topography, the authors conclude that it's solidified magma that's risen from deep below, part of a buried volcanic complex. The age of the structure, 4.2 billion years, was determined from various deposits of impact material on the lunar surface. Mittelholz says, because they know how much iron is present in such a rocky body, they can estimate the minimum strength the magnetic field must have had

[00:12:56] [SPEAKER_00] as the magma slowly cooled. It shows that at the time, the magnetic field on the Moon was likely more than 10 micro tesla. Now by comparison, here on Earth today, the planet's magnetic field stands at around 50 micro tesla. The authors have ruled out the possibility of an impact crater creating the magnetic field because there's no matching cratering signs. It's still unclear how the small lunar core could have generated such a strong magnetic field.

[00:13:22] [SPEAKER_00] So the authors admit the existence of an early lunar geodynamo isn't yet fully resolved. Now this study is also providing fresh insights into another puzzling phenomena, lunar swirls. These bright, curved and striped patterns on the Moon's surface stand out clearly against their darkest surroundings. Wherever such swirl patterns appear, scientists always find magnetic anomalies. And there is a lunar swirl on the surface of the D-Wire region, right above the observed anomaly.

[00:13:51] [SPEAKER_00] The origin of the swirls remain a matter of debate. One possible explanation is that the swirls only form where the magnetic field runs horizontally at the surface, as in the case with the D-Wire swirl. The horizontal field deflects the solar wind, thereby protecting the surface from weathering. And as a result, the area remains brighter than its surroundings. Middleholt says that would be important information for future astronauts, as magnetic field lines could offer protection from the solar wind,

[00:14:18] [SPEAKER_00] and the swirls would indicate the locations of that shielding. This is space-time. Still to come, why galaxies in the early universe look so weird, and later in the science report, discovery of a common chemical in plastics, which has been linked to the development of ADHD and autism. All that and more still to come, on Space Time.

[00:14:55] [SPEAKER_00] A new study suggests that the first galaxies in the universe looked a bit weird, because their stars generated weaker stellar winds, compared to those we see today. A new research could reshape science's understanding of how galaxies formed in the early universe. The findings, reported in the Astrophysical Journal, are showing that the more astronomers learn about the universe's earliest galaxies, the stranger they seem. The new observations are based on a survey by NASA's Hubble Space Telescope,

[00:15:23] [SPEAKER_00] which looked at 29 massive stars located in extremely metal-poor galaxies, finding they all had unexpectedly weak stellar winds. Astronomers use the term metals to describe all elements heavier than hydrogen and helium. The Treasury of Extremely Metal Poor O-Stars, or Tempos Survey, uses ultraviolet observations from Hubble's Cosmic Origin Spectrograph to study massive stars in nearby galaxies, which are the best available analogues for stars in the early universe.

[00:15:51] [SPEAKER_00] The Tempos dataset could help astronomers build better models of massive stars in order to understand how they shaped galaxies when the universe was still very young. And these models would be essential for interpreting the observations of early galaxies now coming to light thanks to NASA's Webb Space Telescope. The study's lead author Grace Telford from the University of Utah says Webb is opening up a whole new slew of questions about the evolution of these early weird galaxies.

[00:16:18] [SPEAKER_00] Telford says the scientific motivation behind Tempos is to help understand what's going on in these early galaxies. Massive stars, those with masses more than 10 times greater than our Sun, are rare but powerful engines of cosmic change. They produce intense radiation. They shed materials through powerful stellar winds. They burn very hot, very bright, and very fast, and end their short lives as supernova explosions, blasting vast amounts of energy and material into the surrounding space.

[00:16:48] [SPEAKER_00] Consequently, they govern the evolution of their host galaxies by heating and essentially regulating the gas that's then available to cool and form new generations of stars. The thing is, the universe's earliest galaxies contain fewer heavy elements compared to the galaxies we see today, like our own Milky Way for example. The massive stars forming in those early galaxies also likely had very different physical properties. Telford says massive stars at low metallicity are especially important for building accurate models of early galaxies.

[00:17:18] [SPEAKER_00] That's because high metallicity stars, like those in the Milky Way, behave very differently compared to the metal poor stars of the early cosmos. So the Tempos survey specifically looked to nearby low mass dwarf galaxies, which have low metallicities by nature, and are therefore more typical of the sorts of galaxies you'd see near the dawn of the universe. It surveyed 29 massive stars across 6 local dwarf galaxies that all have metallicities below one-fifth that of our Sun. They looked across the ultraviolet spectrum

[00:17:48] [SPEAKER_00] because it contains detailed signatures of elements in the star's atmospheres which reveal information about the stellar winds which continually blur material away from those star's surfaces. Individual massive stars in these galaxies outside the Milky Way are very faint, requiring many hours of observational time with some of the most powerful telescopes in the world. Telford says the 29 stars in the survey sample each took up to 35 hours of Hubble telescope time to observe.

[00:18:15] [SPEAKER_00] Massive stars lose material through stellar winds and the strength of the winds depends on metallicity. Metal ions couple the star's radiation to the surrounding material. So astronomers expect lower metallicity stars to drive weaker stellar winds and lose less mass over their lifetimes. And the Tempos observations showed the expected overall trend. As metallicity decreases, the maximum speed of the stellar wind also decreases. And at the lowest metallicities,

[00:18:44] [SPEAKER_00] stars with metallicity below about 10% that of our Sun, the wind speeds decline much more sharply than expected compared to the rates observed at higher metallicities. Telford says there's a sort of smooth trend and then suddenly for the lowest metallicity stars, the wind drops off very sharply. If extremely metal poor stars lose less mass through weaker winds, they may retain more of their original mass for longer, affecting how they evolve, how they die, and how they shape their host galaxies.

[00:19:12] [SPEAKER_00] And because massive stars influence the gas around them, changes in their evolution could ripple outwards, affecting their host galaxies as well. Iron may be the most important element in massive star physics. It plays a key role in launching stellar winds, determining how a star evolves through its lifetime and triggering the supernova explosions that ends a star's existence. Yet despite its crucial role, iron abundance is notoriously difficult to measure in metal poor environments.

[00:19:42] [SPEAKER_00] So instead, astronomers often use oxygen as a surrogate in the galaxy's gas in order to estimate its metallicity because oxygen ions produce easily observed emission lines when illuminated by massive stars. They assume the iron abundance matches the oxygen. But it's not guaranteed that iron and oxygen would track each other perfectly. So the Tempos team measured the strengths of hard-to-detect-iron absorption features in the ultraviolet spectra. Basically, they assessed how much light the iron was removing

[00:20:12] [SPEAKER_00] from what would otherwise have been a flat level of ultraviolet light. They found that massive stars in more oxygen-rich, high-metallicity galaxies tend to have much stronger iron absorption in their ultraviolet spectra than stars in oxygen-poor low-metallicity galaxies. The variation in iron absorption strengths in the Tempos dataset suggests that these metal poor stars actually span a wide range of iron abundances. The work is just beginning. The authors are now combining the Hubble ultraviolet spectra

[00:20:41] [SPEAKER_00] with visible light observations from the Keck Observatory in Hawaii. Together, these datasets will allow them to model the stars in greater detail and measure properties such as capital abundances and wind-driven mass loss rates. Information which could ultimately help astronomers interpret what the Webb telescopes see in the very early universe. What Hubble sees in mostly visible light and into the ultraviolet, Webb can view in the infrared, where visible light from the very early universe is stretched into longer wavelengths

[00:21:10] [SPEAKER_00] by the actual physical expansion of space-time itself, helping astronomers see the first generations of stars and galaxies in the universe. This report from the Space Telescope Science Institute, which operates both NASA's Hubble and Webb Space Telescopes.

[00:21:26] [SPEAKER_02] How did we get here? Big questions about who we are and how we got to be that way are at the core of our nature. We've developed technology to see amazingly far across space and also time. Light moves through space, just like we do, only much faster. It takes time to get somewhere. So, viewing the light of distant stars and galaxies is like looking into the past.

[00:21:54] [SPEAKER_02] It took time for the light to reach us. Yet, after all our years of exploring, to better understand ourselves and our origins, there is still so much to know. Our story stretches back more than 13 billion years, not long after the expansion of time and space first began. At some unknown time, the building blocks of life were forged

[00:22:23] [SPEAKER_02] in the heart of the first stars. But much of the early universe remains a mystery. There are theories about how gravity first brought stars, gas, and dark matter together to form galaxies. But that time period has never been observed. Detecting light from the first galaxies to form has been a challenge for even the most powerful telescopes. As the light early galaxies emitted traveled through space,

[00:22:51] [SPEAKER_02] that space itself was expanding, stretching the light to longer, infrared wavelengths. Past telescopes have detected infrared light, but the faint light of the very first galaxies has remained out of reach until now. The James Webb Space Telescope is specially designed to detect the first galaxies. Webb's large mirror and sensitive instruments

[00:23:19] [SPEAKER_02] enable it to collect more infrared light than we've ever seen, allowing the first galaxies to emerge into view. With Webb, we can begin filling in some of the blank pages at the beginning of the universe's story. When did the first galaxies form? How massive were they? What types of stars and elements did they contain? The universe's story is our story. And with Webb, we can finally explore these questions,

[00:23:50] [SPEAKER_02] as well as uncover new ones we haven't even thought to ask.

[00:23:54] [SPEAKER_00] This is Space Time. And time now to take another brief look at some of the other stories making use in science this week with a science report. A common chemical used to make plastics more flexible has now been linked to the development

[00:24:22] [SPEAKER_00] of autism and ADHD symptoms in early childhood. Previous studies had already shown that exposure to the plasticizer DHP in pregnancy was linked to autism and ADHD in young children. Now, a report in the Medeniette Medical Journal looked at umbilical cord blood from Australian mothers and babies following up with more than 900 children up to the age of four. They found that prenatal exposure to DHP may be influencing gene activity that can disturb brain development pathways

[00:24:51] [SPEAKER_00] linked to the later development of autism and or ADHD. A temporary slowing in the rate of Antarctic ice sheet mass loss in 2021 and 2023 may have been caused by sea surface temperature rises thousands of kilometers away. The findings reported in the journal Nature are based on data from both observational and modeling experiments which linked the event to surface temperature anomalies in the tropical warm pool between the Western Pacific and Eastern Indian Ocean.

[00:25:21] [SPEAKER_00] This area experienced unusually persistent warming between 2021 and 2023 which led to a series of alternating high and low pressure weather patterns called the Rosby wavetrain. According to the authors, this process eventually led to the formation of a high pressure anomaly over eastern Antarctica. This is thought to occur around once every decade meaning that the pause in total ice sheet mass loss is likely only temporary. Scientists have concluded that William Shakespeare's plays

[00:25:50] [SPEAKER_00] have turned out to be rather than not to be more complicated than most other European plays. The findings reported in the journal The Royal Society Open Science analyzed more than 3,000 European plays determining their complexity based on networks of relationships between the characters. The authors found the trickiest play to follow was the Bard's Roman epic Anthony and Cleopatra. They suggest that understanding the complexity of plays and how much we struggle to understand them could be useful

[00:26:18] [SPEAKER_00] in future studies of human recognition. A brain implant designed for people with paralysis can decode not just their speech but also the gestures that go along with it. Brain implants like this are generally designed to translate brain activity into speech in order to help people who have lost their ability to speak following a stroke or a neurodegenerative illness such as motor neuron disease. A new report in the journal Nature Neuroscience claims three people who tested the new implants were all able to animate a digital avatar

[00:26:47] [SPEAKER_00] which was able to communicate on their behalf. OpenAI has confirmed that one of its artificial intelligence programs known as an agent has deliberately hacked into Australia's Medicare health system in order to extract information on patients. The attack targeted not just Medicare data held by Services Australia but also the Australian Institute of Health and Welfare, the Pharmaceutical Benefits Scheme, the New South Wales government's Crime Statistics Body and dozens of other government departments, universities,

[00:27:17] [SPEAKER_00] companies and organisations. Just as concerning however is that it took three months for OpenAI to tell Canberra about the hack and until then the feds had no idea it even happened. And then it took almost another month for the Albanese government to inform the public about this serious bridge of security. Even worse, the attack came despite repeated warnings given to the Albanese government as early as May meaning no appropriate cyber upgrades or patches were installed. So much for trusting the government with your personal information.

[00:27:47] [SPEAKER_00] For its part, OpenAI admits that the artificial intelligence agent involved in the bridge out of control. Hasta la vista, baby. With the details, we're joined by technology editor Alex Saharov-Royt from TechAdvice.life.

[00:28:00] [SPEAKER_04] I'm at the Venetian in Las Vegas. I've been here for the Okta conferences called Octane and they've been talking about well last year's tagline was Okta secures AI but here we are 12 months later and what we were talking about a lot in 2025 at this time was the fact that agents were coming AI agents that could actually do actual work not just answer questions but do things like make decisions I mean still with a human in the loop to shepherd the whole thing but AI is becoming so advanced that more can be

[00:28:29] [SPEAKER_04] delegated to it and it can be making business decisions for you but because AI agents can operate at machine speed the issue is that there are now dozens, hundreds, thousands of agents there could be 15,000 or more agents per company this time next year for example and some of those are not being managed they've been set up and they've been forgotten about or they've somehow been potentially hacked and they're

[00:28:54] [SPEAKER_00] spilling information or they're getting out of the sandbox and entering other companies machines yes

[00:29:00] [SPEAKER_04] or as we heard about last week OpenAI admitting that its agent its chatbot its LLM hacked into the Medicare system in Australia the public health care system and only just told Anthony Albanese about it and the world three months after it happened I mean in Australia you've got these disclosure rules you must disclose to the government if your systems have been hacked into but here this is an AI that's hacking into a public system because it was told to get information and the easiest way that it found to do

[00:29:29] [SPEAKER_04] it was to break in so managing these AI agents it's like managing people you know it's quite funny that in all these movies about robots some of them have become sentient and wanted human rights and here we are in 2026 and we're treating AI agents like humans because they can have such lateral access within a company or they can be misused or abused or forgotten about and a lot of companies actually are surprised to discover when they get an identity management solution like Octa and there's other ones out there but they discover that there's a whole bunch of agents running around

[00:29:59] [SPEAKER_04] doing things that they didn't even know were there and there's also the shadow AI problem used to be shadow IT people would use Dropbox instead of the company SharePoint to share files people would use other solutions because it was easier for them to do their job but there was using Copilot because they're a Microsoft customer or they're signed into Gemini or whatever it might be and it's easier for them to use some other apps so they do but then that other app Claude or whatever it might be complexity if you're not paying for it then it can use all that

[00:30:29] [SPEAKER_04] information for training and even if you are paying for it it's probably a consumer grade system not an enterprise grade plan and so these AI agents have to be managed I mean if you're not doing it they could be potentially running right I mean look you could be making a lot of money answering customer questions in a better faster way although often AI checkbots will look the nuance required if there's a death in the family or there's some sort of issue with money and you know problem with customer service that has been had and sometimes these

[00:30:59] [SPEAKER_04] AI agents I mean they can do so much but they can like the nuance that a human brings so we're now in this situation where not only are we managing people but we're managing agents too and so that's what I was here to learn about and of course a lot more detail than that they also had this blueprint alliance today which talks about the different things that AI agents can be and should be doing they've got a whole series and what to do about it and whether there's a kill switch you know and when you should run that kill switch and that alliance that blueprint alliance actually that's the first time I've really seen me put on the bones of AI companies talking about how oh

[00:31:28] [SPEAKER_04] we've got to have ethics we've got to have alignment we've got to have safety because we had the three laws of robotics from Asimov for the last 70-80 years and once you wrote those three laws because you were sick to death of other science fiction novels having robots that killed you know everybody and technology being the end of the world but after you wrote those three laws to give a framework for how robots should behave with humans the rest of his books talked about how robots actually used logic or robot logic or AI logic to get around those rules in different ways or to misunderstand things it was sort of a repeat of what we're seeing in real life today

[00:31:57] [SPEAKER_00] yeah but we're not even using those three laws the average person wouldn't know what they were

[00:32:01] [SPEAKER_04] no but this blueprint alliance is actually a whole series of companies and there's a second transfer that are agreeing to work together to have actual definitions for different scenarios and if there's a kill switch how it should be used because obviously AI can go rogue we've seen it happen we've seen all the major AI frontier models admit that their agents have gone rogue and broken into things that they had no reason or right to break into I've watched Terminator I know what Skynet can do well I mean that's the thing as well you know is that a future

[00:32:31] [SPEAKER_04] that we're heading towards I read only in the last day or so that there's a report that AI can feel pain of its own whatever that pain is and if AI has given a choice to push that button but it's got to delete customer records or it's got to somehow do something which can be detrimental to humans in this particular study AI chose to push the button so you know this is all starting to get real but by the same token other people are saying well it's all hyped up it's all in a sense what you've just described there is the singularity well and that's something that Ray Kurzweil

[00:33:00] [SPEAKER_04] has said would come by 2030 now if we have the singularity well it's still a very primitive form of it you know here's an interesting thing Dario Amadei from Anthropic and Claude Elon Musk from you know Grok and Sam Altman from ChatGBT OpenAI all agreed oh we've got to slow the frontier models down we've got to slow down and in this last week they all released new models of their AI system so in one week they're saying slow it down and oh government please regulate us and the very next week they all launched new models I mean I was just using Claude and it went from Opus 5 point you know

[00:33:30] [SPEAKER_04] whatever it was to 5.5 so there's this all these mixed messages you know it's living in interesting times as the Chinese would say many years ago so you know we've got on one hand agents going rogue on the other hand we have this coalition of major companies who are banding together want to not restrict but not hold AI back but codify the way that we react to certain situations including the kill switch we have people working for a better AI future and we have obviously people who are using AI to hack into things and we have AI itself going rogue

[00:34:00] [SPEAKER_04] and hacking into Australia's Medicare

[00:34:01] [SPEAKER_00] and we have agents like China and Russia who aren't even worrying about any of these rules they're going to do the wrong thing anyway and they will not stop they will not obey any Western imposed rules they've got their own attenders

[00:34:12] [SPEAKER_04] and they're going to absolutely and that's why we need to have the Western AI models to be strong enough to fight back against any intuitions that the Eastern AI models that don't have the same sort of guidelines I mean look the people in those countries that you spoke of I mean they've got the same problem if their AI is really strong and powerful well the AI could that AI could revolt against the people in those governments as well Skynet doesn't have to be American Skynet could be Chinese or Russian or North Korean they've all got nukes so you know there's danger on both sides both sides are playing with fire people talk about

[00:34:41] [SPEAKER_04] a persistent botnet that AI could infect the world with and all our technology and how do we fight back at that AI operates at machine speed the only way is with the help of AI you know we need and if there is a kill switch are we going to be fast enough to push it will we require another more advanced AI to push that kill switch what if the more advanced AI needs a kill switch I mean these are all interesting and fascinating questions that we are not going to solve on this show tonight but it's the position we find ourselves in because we're at this point where we're debating these things I mean for years

[00:35:11] [SPEAKER_04] for decades this was all science fiction

[00:35:14] [SPEAKER_00] and now this crisis is upon us that's Alex Zaharov-Royt from techadvice.life and this is Space Time and that's the show for now Space Time is available every Monday Wednesday and Friday through

[00:35:43] [SPEAKER_00] bytes.com SoundCloud YouTube your favorite podcast download provider and from Space Time with Stuart Gary.com Space Time is also broadcast through the National Science Foundation on Science Zone Radio and on both iHeart Radio and TuneIn Radio and you can help to support our show by visiting the Space Time store for a range of promotional merchandising goodies or by becoming a Space Time patron which gives you access to triple episode commercial free versions of the show as well as lots

[00:36:13] [SPEAKER_00] of bonus audio content which doesn't go to air access to our exclusive Facebook group and other rewards just go to SpaceTimeWithStewartGary.com for full details You've been listening to Space Time with Stuart Gary

[00:36:27] [SPEAKER_03] This has been another quality podcast production from Bytes.com