Google Cloud Speech-to-Text API (Chirp) + Gemini 2.5 Flash Non-Thinking. Treat it as a searchable index of what was broadcast, not a quotation record.
00:00:00products because of whether they're
00:00:04safe, and in order to have that conversation, we need to have
00:00:07that conversation with the government in the room. the conversation
00:00:11right now is that we are in the precipice of
00:00:15something really scary. is this five
00:00:18alarm fire? i you know, i don't, depends what you
00:00:21mean by five alarm fire. what i would say how would you describe it?
00:00:26kinds of people who are worried, it's a lot of what what i would describe is so something i've been saying?
00:00:32since uh since we started antropic, almost since before we started antropic, is that this
00:00:35technology is on an exponential right, and that's a mathematical
00:00:38function that you know doubles every six months or doubles every three months or
00:00:42something like that, and so when something doubless
00:00:46every small amount of time, it compounds very quickly, right?
00:00:49it's like you have one, then two, then four, then e, then 16, then
00:00:5332,
00:00:55and as I was saying that, what I was trying to say is this
00:00:58technology, is going to jump much more quickly
00:01:02than we imagine, and but even as I was
00:01:05saying that to myself, I don't think I fully just
00:01:09appreciated what it would actually be like when the progress was
00:01:12as fast as it was, so we're we're an exponential curve looks a little
00:01:16like this, where are we on the curve? we're kind of we're kind of on the bend where it's starting to
00:01:20get steeper, so is that five alarm fire to you?
00:01:23what it's what it what it means, it it doesn't mean
00:01:27we need to panic today, it doesn't... we need to
00:01:31shut it all down, what i i would say,
00:01:34it's it's a warning sign, it's a warning sign that we need to
00:01:38slow down, we need to make this technology
00:01:41carefully and we need to make sure that
00:01:44the our safeguards, our ability to
00:01:47understand it, our ability to control it
00:01:50keeps up with the pace
00:01:54at which the technology is
00:01:55happening.
00:02:02Yeah, I think that they're really serious. It's a sign that these
00:02:05people are noticing that the recent incidents
00:02:09um, are much worse than lot of people were expecting, and that we need to do something pretty
00:02:13serious to handle them and make sure that something terrible doesn't happen.
00:02:16Yes, uh, generally speaking, one way of making an
00:02:20AI not do something bad is to put guard rails around it, so that even if it
00:02:23tries, it can't break out of the box or something like that, but the problem
00:02:27is that this kind of approach doesn't scale very well as the... get more and more
00:02:30capable, alignment is about making the AI
00:02:34inherently want to pursue the goals that we wanted to pursue or have the values
00:02:38that we wanted to have, and that's the kind of thing that we need as they get more advanced.
00:02:41yeah, it's there's there's different ways it can go wrong, but at least one of the ways it can
00:02:45go wrong is sort of the like literal
00:02:48genie, you tell it to do something, but you slightly
00:02:51misspecify what you wanted to do and it gets way out of hand. i think the hugging
00:02:55pase incident is an interesting example of this where the ais were sort of
00:02:59trained to try and... get a good score and they found out that the best way to get the good
00:03:03score involved breaking out onto the internet and hacking a
00:03:06public company. i think it's a hard question. um,
00:03:09so definitely, first of all, amusingly, open air was founded in a
00:03:13similar way where sort of sam and Elon Musk were like, we're not sure that we trust Google to
00:03:17be making this stuff, so actually all of the top
00:03:20missing under the rubble and competitors were born out of not trusting each other. i think that there is definitely some component of
00:03:24them spinning this news in a way that is more favorable to them, but very
00:03:27fundamentally if they could snap their... fingers and have it be that their
00:03:31ais do not break out and randomly hack things, they would definitely want to do
00:03:35that, so I don't think it's all spin, I don't think it's all someelaborate con, I
00:03:38think they are genuinely scared, but you do have to take the
00:03:42exact phrasing with a pinch of salt, the honest answer, and this is kind of
00:03:45unfortunate is that AI is becoming more and more of a
00:03:49critical piece of national security infrastructure, and
00:03:52so I would guess that the US government had at least some part to
00:03:56play in the decision, the more it
00:03:59becomes the case that... things of you, for example, previously anthropic and open
00:04:02ai had their own internal programs to figure out who would get private access to these
00:04:06models for defense, now the US government has kind of taken over that. i
00:04:10think UKAC is absolutely wonderful, i think that some part of what's
00:04:13going on is the the UK government doesn't have as much leverage and the
00:04:16third party auditors they're looking for are very much us auditors. very
00:04:20difficult question, um,
00:04:22mean a couple of things, what have we seen lately? so i think one thing that people
00:04:26underrate is that the
00:04:28deal. and cyclicals are one of china is in some ways more sympathetic to slowing down,
00:04:31because they're aware of the fact that if AI gets very powerful, it could be bad for
00:04:35the PRC, could disrupt their ability to sort
00:04:38of um function effectively as a sort of dominant
00:04:41government. um, so actually compared to the
00:04:44US, they have some appotype for it, they are also currently behind,
00:04:48so there's a lot of specifics in how you do anything like slowing down or
00:04:52restraint, but there are definitely versions of it which i would guess will be pretty appealing to
00:04:55them, also that yeah,
00:04:58so they're not exactly. open source in that what they release
00:05:02is the contents to run them yourself but not to actually build them yourself or see how they've
00:05:05done see how they've been built like traditional open source software you can see all the
00:05:09code youif it's safe but decision to to
00:05:13public things, it's very common in tech in general for people who are a
00:05:17little behind the frontier to want to make their things more available, so sort of google doing this with
00:05:21android was a similar story and windows doing it compared to apple by
00:05:24releasing sort of windows OS. um, i'm certainly
00:05:28hoping that it ends up being a dam squib, but i think that the chances that it
00:05:32aren't are substantial enough that we need to take this very very
00:05:35seriously. mean, if you look at the
00:05:38the sort of most heavyweight, well respected.
00:05:42AI research luminaries um and the predictions that the people who
00:05:46are most worried were making 10 years ago are beginning to come true, so
00:05:50definitely watch this space,
00:06:10society of leaders has.
00:06:23Before the representatives of the European Parliament in Strasbourg, the President of the European Commission
00:06:27Fonderline summarized the state of the union. German Chancellor
00:06:30Merce is record unpopular, the costs of the war in
00:06:34Iran are growing, and this morning there are reports of an attempted Houthi attack on
00:06:37the holiest place of Islam in Mecca.
00:06:41In Gaza, a new tragedy happened. Welcome to
00:06:44Global.
00:06:51European Commission President Ursula von der Leyen delivered her annual
00:06:55State of the European Union address in Strasbourg, warning
00:06:58that the Union is facing its most insecure position
00:07:01yet due to aggressive Russia, economic pressures and hybrid
00:07:05threats. As key priorities, she listed launching
00:07:08the European economy, reforming the energy market and strengthening
00:07:11security protocols similar to NATO. In the context of global
00:07:15politics, she stated that isolationism is not an option and
00:07:18announced the strengthening of alliances, proposing that
00:07:21Canada become the first associated member of the European
00:07:24Union. But above all, dear
00:07:27Mark,
00:07:30Europe and Canada believe in
00:07:31democracy.
00:07:34We believe that power does not belong to the strongest,
00:07:37the richest, the loudest, but to all of us. Democracies
00:07:41can choose with whom
00:07:42to cooperate.
00:07:47That is why we are voluntarily joining
00:07:49forces.
00:07:52After that, we move on to the alliance for the future, our goal is
00:07:55to create a common space for prosperity and economic
00:07:59security. We will work on intelligent manufacturing,
00:08:02we will establish a technological alliance, we will integrate defense
00:08:05industrial bases. The Arctic will become our main
00:08:09common project, we will cooperate in the field
00:08:12of energy, critical minerals, batteries, artificial
00:08:15intelligence, quantum technology, cybersecurity and
00:08:18economic security. Our partnership is not directed against
00:08:22anyone, but strives for strength through
00:08:25unity. We want to strengthen
00:08:27our relationship with Canada as much as
00:08:30possible.
00:08:33Dear Mark, I said we urgently need to redefine
00:08:36our partnership.
00:08:39That is why I would like us to open
00:08:43the door for Canada to become the
00:08:46first associated
00:08:47Union.
00:08:51I was horrified by the images of last week's funeral of
00:08:54Ratko Mladić in Belgrade.
00:09:15He committed the cruelest crimes against
00:09:18humanity,
00:09:21there is simply no place in the European Union for glorifying war criminals.
00:09:35All victims of the wars in the former Yugoslavia deserve to be remembered
00:09:39and all their families deserve
00:09:42our sympathy,
00:09:46because that is what we as a union stand for, for truth,
00:09:50justice, respect of
00:09:52our values.
00:09:59A lot was said, but little concrete, said
00:10:02foreign policy analyst Mirjana Rakić to our new
00:10:06day.
00:10:08She used "we need" a lot.
00:10:12"we will, but how, in what time, what are those
00:10:16deadlines, some new alliances are being formed, an alliance for
00:10:20electricity, an alliance for artificial
00:10:23intelligence, The most interesting to me seems to be this
00:10:26European security alliance where in a way
00:10:30a counterbalance to NATO is being
00:10:32created.
00:10:41German Chancellor Friedrich Merz canceled his trip to the UN General Assembly
00:10:45in New York. He remains in Berlin for key party
00:10:48meetings. The decision comes at a time of great pressure after the heavy
00:10:52defeat of the Conservatives in Saxony-Anhalt.
00:10:55New polls show that the right-wing Alternative for
00:10:58Germany is leading nationally with a
00:11:01record 30% support, while the CDU
00:11:04CSU bloc fell to 20%. The upcoming state elections in Berlin
00:11:08and Mecklenburg-Western Pomerania on September 20 will be a new
00:11:12test for Merz, while there is already speculation in the public about
00:11:16a possible change in government leadership. My
00:11:19current interest is in the substantive issues of how we can move forward.
00:11:22We are an elected government. We are elected for four years.
00:11:25I am elected for four years, and I want us to use these
00:11:28four years to lead the country out of this difficult phase.
00:11:32phase.
00:11:41The American war against Iran has so far cost $38 billion
00:11:45with estimates that monthly costs will increase by $3 billion,
00:11:48the Congressional Budget Office announced. The conflict has
00:11:52exhausted ammunition supplies and caused a rise in energy
00:11:55prices, which could increase inflation by
00:11:580.5% in the first three months of next year, while the White
00:12:02House and Pentagon claim that the military has enough resources for all
00:12:05objectives, the US House of Representatives has for the third time voted
00:12:09for a resolution to end the conflict. The decision comes before the midterm
00:12:12elections, and voters are not happy about the jump in fuel prices
00:12:16caused by the blockade of the Strait of
00:12:18Hormuz.
00:12:26Saudi Arabian air defense intercepted a Houthi drone near
00:12:29Mecca, which is the first attempted attack on this city since
00:12:322017. Due to security risks, the US Embassy
00:12:36warned its citizens to avoid the border area with
00:12:39Yemen, due to damage to the pipeline to the Red Sea in
00:12:42drone attacks, Saudi Arabia reduced oil supplies to
00:12:46Europe, which caused a jump in oil prices, so European companies
00:12:50are forced to look for alternative sources
00:12:52of supply. In
00:12:58The Yemeni armed forces will continue to protect the airspace of our
00:13:01country and preserve its sovereignty from any aggression or
00:13:04violation by all available means, and the sky of Yemen will not be
00:13:08violated, God
00:13:09willing.
00:13:16At least 20 Palestinians, including five children, were killed
00:13:20and 25 were injured in the collapse of a six-story building in
00:13:23Gaza City. The building was damaged in Israeli attacks
00:13:26and served as a shelter for more than 60 displaced people.
00:13:30Rescue teams are trying to reach those trapped
00:13:33under the rubble, but their work is hindered by the lack of heavy
00:13:37machinery. Authorities in Gaza state that about
00:13:408,500 people across the strip are still
00:13:45call on the international community for urgent intervention and an end to
00:13:48the blockade.
00:13:59to support uh to support these families uh
00:14:01situation as you see is tragic because we don't have the equipment and
00:14:05machinery that is needed so there are actually
00:14:08people that are digging uh with the bare
00:14:11ends we need to
00:14:14prepare uh there are approximately 400
00:14:17buildings at risk of collapse and we know that these
00:14:20buildings will collapse especially now with the upcoming
00:14:24winter. uh we need
00:14:27support uh most importantly with
00:14:30machinery, equipment, among others also
00:14:33to our colleagues here that don't have oil
00:14:37and spare parts uh to continue and to work with all the
00:14:40equipment that is still in Gaza, we are running
00:14:44against time, this is a tragedy within a larger
00:14:48tragedy, and it is important
00:14:52that we're all coming together as international community.
00:15:04That's all for this edition of Global, the latest world news is, as always, on
00:15:08our web portals and social networks. Goodbye until tomorrow.
00:15:44Let's take a closer look at some of the biggest stories in tech right now:
00:15:47could quantum computers break the internet as we know it? Meta's AI
00:15:51is checking your age, but at what cost? Nvidia wants your
00:15:55pc to think and act for you, and AI is even stepping
00:15:58into religion. Let's break it
00:16:00down.
00:16:03Does the internet have an expiration date? Google says by
00:16:062029, digital world could change completely, the
00:16:10reason q-day. That's what experts called the moment when quantum
00:16:14computers are able to crack the encryption that keeps the internet
00:16:17safe. Think of a mase for second. Normal computers try
00:16:20one path a time, while quantum computers explore all paths
00:16:24at once. That capability makes them very fast. In May, a
00:16:28Chinese quantum computer reportedly set a new speed record in a
00:16:31particular task. It completed an extremely complicated calculation
00:16:35in 25 microseconds. The world's... faster supercomputer would need a
00:16:39treated cillion years to solve the problem, that's a one with
00:16:4242 zeros. please note, this benchmark was reached in a test
00:16:46setting and does not mean quantum computers can already crack today's
00:16:49encryptions, but it shows the technology is evolving a
00:16:52breath-taking pace. what does that mean for cyber security? who's at
00:16:56risk? if quantum computers were able to break current
00:16:59encryption, lot would be exposed. first, national
00:17:02security and infrastructure. Think of power grids, transport systems
00:17:06or military. communication: with a quantum computer, bad actors could
00:17:10in theory weaken a country's vital infrastructure. second, the
00:17:13global economy, banking and crypto heavily rely on
00:17:16encryption, a single hack could cost trillions of dollars, and
00:17:20finally, you, every private dm, bank password and
00:17:23medical record encrypted at the current standard would be vulnerable.
00:17:27sounds scary, but how real is this threat right now?
00:17:31Okay, let's not panic too much for now. cyber security companies
00:17:35love talking about Q day because fear helps them. new quantum safe
00:17:38encryption. Apart from that, Google has published a timeline for making its
00:17:42operating system Android quantum resistant, promoting themselves
00:17:46as a trustworthy player for the future, but the threat is
00:17:49real, and we have already reached the preliminary stage. Hackers are
00:17:53stealing encrypted data, even if they can't read it yet, they storing it,
00:17:56waiting for quantum machines to catch up. When that happens, old data could become
00:18:00readable overnight. The phenomenon is already so wide spread
00:18:04that there's a term for it, harvest now, decrypt later. or
00:18:07h&DL in short, so are we just doomed to face a digital
00:18:11apocalypse? not exactly, there's a global push to
00:18:15switch to PQC (post quantum cryptography). new encryption
00:18:18designed to resist quantum attacks. these are ultra complex mathematical
00:18:22riddles that even quantum computers can't easily untangle, and the push
00:18:26is not just coming from the private sector. governments are starting to set
00:18:29deadlines and restrict software that won't hold up in a quantum
00:18:33future.
00:18:35meta's AI might soon be..." analyzing your photos and videos to
00:18:39estimate your age. The new AI system looks at visual clues like bone
00:18:43structure to flag undererage users. The announcement comes after
00:18:47Mata was hit with a $375 million dollar
00:18:49fine in March for misleading people over child safety. The
00:18:53first rollout is coming to the US, Facebook only as well as
00:18:57Brazil and the EU. Meta claims this AI is more reliable
00:19:00than past age checks, which kids often hack with fake birthdays or
00:19:04even fake mustaches. Is matter's move a breakthrough for? child
00:19:07safety or a massive overreach in biometric
00:19:10surveillance. For years, tech giants have struggled to
00:19:13verify H without demanding a government issued ID. Meta
00:19:17Solution, its own AI powered age assurance, similar to
00:19:20systems already on the market. Meta says it has spent over decade
00:19:24building tools to protect teens, but regulators aren't convinced.
00:19:28In late April, the EU ruled that Mata failed to stop children under
00:19:3113 from using its platforms. Data from Australia's new under
00:19:3516 social media band shows. age verification systems don't
00:19:39work well and enforcing a ban is difficult. plus, meta's
00:19:43new measures have privacy advocates alarmed. even if
00:19:46it's only for age checks, scanning a chart's physical appearance
00:19:49raises big questions. where's this data stored? can it be hacked? apart from
00:19:53that, there's no information on how reliable the technology really
00:19:57is. meta insists, this isn't facial recognition
00:20:00and that individuals are not being identified. the AI scans your
00:20:04photos and videos for visual physical science of puberty. it compares your
00:20:08height to your surroundings, checks your shoulder width and facial bone
00:20:12structure that changes as you age, then it cross references this with
00:20:15data from your page. did you post about this suite 16? are your
00:20:19friends all eighth grade? If the system finds that things don't add up, your
00:20:22account is switch to teen mode, limited dms, restricted
00:20:26content, parental controls enabled. to unlock it, users
00:20:29must prove their age with an id or a video selfie to be verified
00:20:33by a third-party tech company. is there a better
00:20:36way? service: like Yoti or Verif estimate H from a
00:20:40selfie without identifying the person, they only pass on the result
00:20:43to the platform. Still, critics worry about data security and what
00:20:47third party providers might do with that data. then there's zero
00:20:51knowledge proofs or ZKPs. The EU is
00:20:54already betting big on this, they launch a specialized age
00:20:57verification blueprint in April 2026, as part of the
00:21:01new digital identity wallet rollout. ZKP's can confirm
00:21:04things like over 18 without sharing personal details, but they
00:21:08only work properly with a massive digital infrastructure that many EU
00:21:12countries are still building. Plus, you still need to trust a government issued
00:21:15app. There's also a bigger debate over who should be responsible for age
00:21:19verification. verification meta says app stores should handle it, so users only
00:21:23need to verify their age once. experts say age verification
00:21:27cannot fix the problem anyway, because the real issue is not who's on the
00:21:30app, but how the app is built. with algorithms designed to be
00:21:34addictive for adults too, and content that isn't safe for 14 year
00:21:37olds, age scanning AI cannot be the fix. and hate,
00:21:41extremism and other harmful content don't just affect kids. changing
00:21:44the platforms would be the better solution. what's your view on
00:21:48this?
00:21:50what if your next pc didn't just respond to inputs, but actually
00:21:54did your work for you. nvidia says pcs will no longer be a
00:21:57tool and instead start acting like assistant. the us tech
00:22:01giant unveiled new breed of AI computers powered by a chip called
00:22:05RTX spark. described as the most efficient pc chip ever
00:22:08built, it's designed to run AI agents locally, no cloud
00:22:12required. nvidia claims this tag will eventually replace the mouse and
00:22:16keyboard by letting ai agents navigate your pc for you, but
00:22:20what exactly is an AI agent and can Nvidia really reshape
00:22:24how we use computers? for the average user, Nvidia's new AI
00:22:27PCs are promising to make life much easier. you just tell the
00:22:31computer what it should do, for example, find the cheapest
00:22:34flight, book it and put the details in my calendar. an AI agent then
00:22:38does this by navigating the computer its own, no more manually jumping
00:22:42between tabs, copying data and clicking through menus. in theory,
00:22:45agents can handle everything you're doing on your computer, even editing
00:22:49files, software. we recently tested an ar agent called open
00:22:53claw. we connected it to a telegram chat and sent it a voice message.
00:22:57the agent said it couldn't process audio, so we simply told it
00:23:00to fix that. it searched the web, installed new software and taught
00:23:04itself how to transcribe audio. nvidia bats it, this kind
00:23:07of automation will be the new normal for personal computers.
00:23:11to make this happen, the tech giant is moving into the cpu market and
00:23:15is partnering with major brands to challenge intel and AMD.
00:23:19for decades, windows C's relied on separate CPUs and
00:23:22GPUs. Nvidia's Rtx Park combines both into a single
00:23:25chip, promising faster performance because data does not need to be sent back and
00:23:29forth between CPU and GPU, and RTX Park
00:23:33is optimized for the whole NVIDIA ecosystem, which already sits at
00:23:37the center of the global AI industry. With that power, Nvidia
00:23:40wants to move AI workloads off the cloud and onto your computer. That
00:23:44would mean faster response times, better privacy and independence
00:23:48from internet connectivity, but It doesn't come cheap, experts estimate
00:23:52the price for an AIPC at this level to be around
00:23:54$4,500 US and autonomous agents brings serious
00:23:58vulnerabilities. AI agents act on their own and sometimes
00:24:02make mistakes, and because they have access to files, accounts and
00:24:06applications, those mistakes can become serious. In a recent
00:24:09case, an AI agent reportedly began deleting emails after losing
00:24:13track of key instructions. It only stopped when the user manually
00:24:17intervened. There are also prompt injections and systems
00:24:20can't reliably distinguish between helpful commands and harmful
00:24:24content. hackers can conceal hidden instructions for your AI agent
00:24:27in an email or a website. some predictions say over
00:24:3140% of agentic AI projects will be canceled by
00:24:342027 due to increasing costs or security risks.
00:24:37nvidia says it's working with Microsoft to deeply embed AI
00:24:41controls into Windows software. the operating system monitors what
00:24:44agents do, flex risky actions and asks for your approval for critical
00:24:48steps. but do we actually want AI to act on our behalf or are
00:24:52we giving away too much control? let us know what you think in the
00:24:55comments.
00:24:57can AI bring you closer to your god? AI jesus and the buddhist
00:25:01AI monk went viral and then in May the pope waid in on
00:25:05the rice of AI in the world. popelio the 14th released
00:25:08magnifica humanitas, his very first official paper and
00:25:11cyclical. the text focuses on how to protect humanity in the age
00:25:15of AI. that's a big
00:25:18the highest forms of in the Catholic Church, but the story goes
00:25:22beyond the Vatican: religion is more and more confronted by AI and faith
00:25:26leaders are debating how to respond. communities around the world are trying to
00:25:30figure out what happens when AI takes over matters of faith. let's break it down.
00:25:34what does the pope actually say? in magnifica
00:25:37humanitas, pope Leo wants about AI's risks. some of these
00:25:40concerns might sound familiar: people losing jobs to automation,
00:25:44AI powered weapons, kids being exposed to dangerous or
00:25:47manipulative AI content. the pope stresses one central idea
00:25:50humans must stay in control of AI and here's something
00:25:54interesting: big teg was part of this conversation andthroppy co-founder
00:25:58Christopher spoke at the post presentation of the text. it's obvious:
00:26:02tech companies no religion still has huge influence on society and
00:26:06faith leaders no AI is already changing how people think, learn and
00:26:10even worship. so what does that actually look like in real
00:26:13life? for Christians, there's an AI version of Jesus, a startup
00:26:17in the United States office paid video calls with AI Jesus trained on
00:26:21the Bible and religious sermons. in Japan, there's an AI robot mon called
00:26:25Buddharoid, it's trained on Buddhist scriptures and can answer questions inside
00:26:28the temple in Kyoto. In Latin America, some church organization
00:26:32are openly encouraging the use of AI for education, sermons
00:26:36and community work, and in India and Pakistan, religious
00:26:39organizations are using AI chat assistance to reach younger
00:26:42audiences online. This is becoming a global trend. Analys say
00:26:46AI based spiritual and worldness apps are now a multi-billion dollar
00:26:50industry. In Muslim communities, the debate is especially sensitive.
00:26:53Some scholars question whether AI can give religious guidance at all,
00:26:57raising concerns about authority, and even whether certain forms of AI
00:27:01might be religiously. missible and that's just one of many problems.
00:27:05AI can sound very convincing, even when it's
00:27:08wrong. so what happens if someone asks in AI for religious
00:27:11advice and gets misleading or harmful guidance, who's
00:27:14responsible? there's also the issue of commercialization.
00:27:18spiritual guidance that was once free or community-based can suddenly come with
00:27:22a price tag like paying for access to an AI Jesus. for some critics
00:27:26that echos all debates about the commercialization of faith, and then
00:27:29there's bias. AI models learn from existing... human data, which
00:27:33means they can inherit sexism, ethnic bias, case
00:27:36discrimination or extremist interpretations, and sometimes the
00:27:40AI presents those ideas like they're the objective truth. so what does
00:27:44the pope suggest? the pope says we need safeguards,
00:27:47things like protecting workers displaced by AI, keeping humans
00:27:51in charge of critical decisions, teaching students how to use AI
00:27:55responsibly and making sure governments actually regulate this
00:27:58technology, not just leaving AI companies to regulate themselves. That's
00:28:02it for me today, bye-bye and see you next time.