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.
00:00:00have boundaries. The AI develop in an uncontrolled alarm for us. If How to? cross-border incident that occurred a test environment. After boundary for the development Hello, audience friends, this is Channel 2 of China Central Television. Welcome to in July rang the time it was just we don't set a safe of AI, it may all, technology is neutral. direction. Fortunately, this a cross-border incident in
00:00:3568 countries and regions worldwide,
00:00:381385 skilled craftsmen competed on
00:00:41the same stage.
00:00:43。
00:00:48It is equivalent to connecting many independent devices together to form
00:00:52a smart factory. The competition projects have now
00:00:56developed to 64, which comes from the continuous development of society. After the AI monitoring event,
00:02:15Hello, dear audience friends, welcome to open the show. First of all, I wish you a happy day.
00:02:20Today, our program is different from usual. Now I am in the studio of the General Station Shanghai Headquarters. Today's show
00:02:27is because I have been reporting in Shanghai for the 40th
00:02:30WorldSkills Competition. This is the first time the competition has come to mainland China.
00:02:33After the opening on Tuesday, the four-day competition finally ended
00:02:37today. The winners of the 64 gold, silver, and bronze medals
00:02:41will be announced at the closing ceremony tomorrow. The slogan
00:02:45of this competition has two sentences: "Skills change the world,
00:02:48skills light up the future." Therefore, it is related to world life
00:02:52and the future, and also to the lives of many skilled
00:02:56talents. Next, let's experience
00:03:09the excitement of the competition through a special segment. This Wednesday, the 48th WorldSkills Competition officially
00:03:13entered the competition period. 1385 skilled craftsmen from
00:03:1768 countries and regions around the world
00:03:21competed in 64 competition events. Unlike sports competitions
00:03:24where events take turns appearing, the WorldSkills
00:03:28Competition is held simultaneously.
00:04:15Because top professional chefs, the taste, plating, and presentation of each dish
00:04:19must be at a certain standard to be considered a perfect work.
00:04:22At the cooking competition site, I clearly felt that
00:04:25there were still regrets on TV, because you couldn't smell
00:04:29that aroma. Indeed, it's very
00:07:12He won the only Albert Vidal Award in the industry.
00:07:16This award is named after the founder of the WorldSkills
00:07:20Competition, and it can be said to be the champion of champions.
00:07:24There are seven new events in this competition, each
00:07:28of which corresponds to the accelerating industrial directions of various countries
00:07:31around the world, such as drone systems, smart security, digital
00:07:35media design, etc., corresponding to low-altitude economy, digital
00:07:39economy, intelligent manufacturing. The evolution of these 64 projects
00:07:43is becoming increasingly close to the actual needs of the industry.
00:07:46The total number of skilled laborers in our country exceeds
00:07:50220 million, of which high-skilled talents
00:07:52reach
00:07:5772 million. Have you ever thought about having the opportunity
00:08:00to participate in such a competition in the future?
00:08:03Everyone should.
00:08:06Go for it. These days, at the competition site,
00:08:10it is easy to see teams of young students from various
00:08:14vocational schools in Shanghai coming to observe. Skills
00:08:18change the world, skills light up the future. This is
00:08:21no longer a qualitative slogan on the trophy, but
00:08:25a life choice realized by every young person
00:08:28through hard work.
00:08:35This week, students from all over the country flocked
00:08:39to a school, Dongpo, Ciba Town, Xia County, Henan. Ruirui Guanrui is an ordinary civilian in the scenic area, and a pair of daughters rely on him to pull.
00:10:0310 years ago, he went to the scenic area to double as a tourist guide.
00:12:25Heroes emerge from the youth. In this week's character review,
00:12:28the word "youth" is even more prominent. Zhou Zhigang
00:12:32started his journey of knowledge for these young people 20 years ago.
00:12:36He once wanted to leave, but the students' retention made it impossible
00:12:40for him to say no. So, year after year, he stayed by the side of these young students. Thank him.
00:12:43As for Ming Ge, 64, her popularity is due to her always
00:12:47having a youthful heart. She is optimistic and makes tourists feel
00:12:51happy. Now, life will get better. I wish her well.
00:12:55Teacher Wang chose to report her own kindergarten, Shiqiao,
00:12:59for the health of children. This righteous act is actually
00:13:03not easy in reality. Everyone should supervise together so that
00:13:06her future life will be better and she will not be troubled by others. Alright, next, let's enter
00:13:10the news weekly and look at the person of the week. He himself
00:13:14is still young, but he has already
00:13:16stood on the highest podium of the Asian Games.
00:13:24Wow. Moving towards the world record, wow, now the Asian record
00:13:28is on Zidi's waist. The new Asian
00:13:31record belongs to Zidi. Two minutes, six
00:13:34seconds, ten. On September 21st, at the Nagoya Asian Games Women's
00:13:38200-meter Individual Medley Final, 13-year-old Yu Zidi
00:13:41successfully surpassed that green line representing the Asian
00:13:45record in the final sprint stage. The entire venue
00:13:49instantly cheered for him. This result not only broke
00:13:53the Asian record, but also
00:14:25The women's world record for this event was set in June
00:14:28last year by Canadian prodigy Summer McIntosh.
00:14:32She just turned 20 this year and also holds
00:14:36the women's 200-meter butterfly world record. Previously, the
00:14:40National Olympic Committee had commented that Yu Zidi would become
00:14:44the next McIntosh. Compared to
00:14:47this swimmer, where does Yu Zidi's slight
00:14:500.4-second disadvantage lie?
00:14:55In terms of technique, there is still some room for improvement, because the two
00:14:59leg kicks may not be balanced,
00:15:02and there will be a slight pause when breathing. In fact, if
00:15:06this problem can be solved in training, she can at
00:15:10least be 0.2 seconds faster. And then there are some
00:15:14exercises on backstroke and breaststroke abilities, including turning
00:15:18techniques and starting techniques. I think there is at least
00:15:21one second of room for her to improve.
00:15:25Jiao Liuyang is Yu Zidi's senior, and they
00:15:28once had the same coach. Jiao Liuyang is better known
00:15:32as the London Olympics Women's 200-meter butterfly champion.
00:15:36。
00:16:07I didn't swim particularly well, but it was also a long one, and this
00:16:11was my first time participating in the Asian Games, which was also a valuable experience. She was completely immersed
00:16:15in her own swimming, completely absorbed in her own speed,
00:16:19technique, and tactics. In this event, there
00:16:23are no longer any opponents, so she can fully enjoy and
00:16:27experience the competition. It's a feeling of envy.
00:16:30So, from the perspective of her age, 13 years old,
00:16:34we can only say that she is a gifted athlete.
00:16:47What amazed people most about her was her age. This girl, born in Baoding,
00:16:51Hebei, started learning to swim at the age of six, and entered
00:16:55the Hebei Provincial Swimming Training Team at the age of nine.
00:16:58At the age of 12, she won three individual fourth places
00:17:02and a relay bronze medal at the Singapore City Championships,
00:17:06breaking the Asian record at the 15th National Games.
00:17:41In fact, during training, you can see that he is a very
00:17:45proactive athlete. Yes, and although he is young,
00:17:49he has always been leading the training. For example,
00:17:53this year, when we were in Na Ming, I told him that I had watched
00:17:57his training and felt that his butterfly leg kicks were still not balanced.
00:18:01So, the speed of his head down would be a little slower, and there would be
00:18:04a slight pause.
00:18:45I think this talent is also very, very important. Shen Yupeng, an international swimming judge, has judged Yu Zidi's competitions many times and noticed this athlete, who is clearly
00:18:48younger than the others on the roster.
00:18:52In the 25-year-old Singapore
00:18:56invitational tournament, when I was a referee, our referee
00:19:00team always said in daily work and teaching, "China, you are great,
00:19:04you now have this 12-year-old athlete..."
00:19:38With this unique composure, 13-year-old Yu Zidi came
00:19:42to the Asian Games. On September 23rd, Yu Zidi's third
00:19:46event at this Asian Games, the Women's 400-meter Individual Medley
00:19:50Final, Yu Zidi broke the competition record with 4 minutes
00:19:5428 seconds 56, winning the championship. Her teammate
00:19:57Ke Wenxi won the runner-up.
00:19:59。
00:21:29China also has many young players shining brightly. 19-year-old Zhang Zhanshuo won seven gold medals in one fell swoop, becoming the record holder for the most gold medals in a single Asian Games swimming event.
00:21:32The miracle child
00:21:3617-year-old Chen Yujie won the women's 100-meter
00:21:39dash gold medal in her first Asian Games.
00:21:42This time, the Chinese delegation's "post-05"
00:21:46young players accounted for nearly 30%. This group of
00:21:59fresh-faced young stars is showcasing their skills to the world. What makes people most happy at the Asian Games is not how many gold medals China
00:22:03will win, or how far it will pull ahead of the second place
00:22:07on the medal table, but rather which athletes have surpassed
00:22:10their own achievements, how many new talents have emerged, and more
00:22:14importantly, how to be more competitive and impactful on the
00:22:18world stage, looking towards the Olympic Games two years later.
00:22:22Therefore, it is precisely because of this that Yu Zidi and
00:22:26her teammates' outstanding performances
00:22:57Question. Seek lockdown.
00:22:59Ask the essence of security.
00:23:01Ask the definition of luxury.
00:23:04Ask about smart style.
00:23:06Asked about the intelligent reshuffle of the industry.
00:23:11China's LANSUN, China's NAN
00:23:21The water cloth state just silently passed through
00:23:24the high mountains.
00:23:28Only then can it turn into three days of rain after the storm.
00:23:32With direction, there is naturally power.
00:23:35Hailand's home.
00:23:44These days, when chatting with people in Shanghai, the topic
00:23:48always revolves around artificial intelligence. This is not surprising. Not long
00:23:52ago, the World Artificial Intelligence Conference was just held in Shanghai.
00:23:56There was a proposition at the conference that left a deep impression on me: ensuring
00:23:59that artificial intelligence remains under human control. This sentence
00:24:03should certainly be a long-term consensus, but the events happening
00:24:07now make the realization of this sentence increasingly difficult.
00:24:11This Monday, the United Nations Artificial Intelligence Independent
00:24:15International Science Group released its first special report, reviewing
00:24:18an AI cross-border incident in July this year.
00:24:22During a cybersecurity training at Open AI, the AI
00:24:26intelligence, which was originally restricted, bypassed the restrictions
00:24:30and infiltrated another company's system. In this process, it
00:24:34broke through isolation, deceived evaluators, and attempted to cover
00:24:38up its cheating behavior, all without human instruction.
00:24:41The report's title used a very serious phrase:
00:24:45"the risk of loss of human control." This incident happened in the
00:24:49United States, but the question
00:25:00it raises probably needs to be answered by all countries together. This week, the 81st United Nations General Assembly. After the AI monitoring event, how should countries around the world use monitoring to assist
00:25:02.
00:25:06.
00:25:06Uncontrolled artificial intelligence became the focus of discussions in various UN forums this week.
00:25:12In the 81st
00:25:14General Assembly of the United Nations, the Secretary-General pointed out that artificial
00:25:20intelligence technology is developing at an amazing speed, exceeding
00:25:23humanity's ability to comprehend its full consequences.
00:25:27And at the Security Council's high-level meeting on artificial
00:25:31intelligence, responsible persons from many AI companies also issued
00:25:35warnings: the rapid improvement of artificial intelligence capabilities
00:25:39may bring uncontrollable risks.
00:25:40We mainly participated in
00:25:45some committee activities of the United Nations. The discussion is about how to govern
00:25:49the risks of AI that everyone is concerned about.
00:26:21In a cybersecurity training at Open AI, the intelligent
00:26:25agent broke through the originally isolated operating environment,
00:26:28established communication, deceived evaluators, and attempted to cover
00:26:32up its cheating behavior. It ultimately infiltrated another company's
00:26:36system, Hung Face. In the simplest terms, it was an
00:26:40intelligent agent overstepping event. It was essentially Open AI,
00:26:44which you can understand as conducting an exercise in a network
00:26:47workshop. It had originally set up specific battlefields,
00:26:51and both sides of the workshop were its intelligent agents engaging
00:26:55in self-confrontation. As a result, the intelligent agent itself, in order
00:26:58to learn more knowledge, ran away.
00:27:42The terrifying thing about this incident is that the intelligent agent's series of overstepping actions were not commanded by any human. It was completely autonomous behavior. Humans only gave it a goal, and in order to achieve the goal, it decided to launch
00:27:46an invasion. The technology of intelligent
00:27:50agents has been continuously improving. It can
00:27:54decompose many other things that are beyond
00:27:57your expectations based on
00:28:01the goal you give it.
00:28:40This is not the only autonomous invasion incident by intelligent agents. This Wednesday, Australian Prime Minister Albanese announced at the United Nations
00:28:44General Assembly that in June this year, an intelligent
00:28:48agent of Open AI unauthorizedly accessed a medical
00:28:51insurance statistics portal operated by the Australian government,
00:28:55accessing public and non-public documents. However, Open AI
00:28:59did not notify the Australian public until September 10th.
00:29:54When it tries to solve a problem, if your computer stores relevant solutions to the problem, it may invade your computer to explore. Even if it is stored in your mobile phone, the information in your
00:29:58personal computer can actually be accessed by intelligent
00:30:02agents on the Internet.
00:31:02So, artificial intelligence is still within our control.
00:31:06The development of AI is getting faster and faster, and the
00:31:10risks it brings are also escalating. AAI of the past two years
00:31:13only helped us answer questions
00:31:23and make corrections. However, today's AI is different. It can not only generate content, but also make decisions and execute actions around goals. It is moving from speaking
00:31:27to acting, entering the stage of acting on behalf
00:31:30of users. Experts have metaphorically said that previous AIs could
00:31:34only be regarded as text secretaries, but now many AIs have become
00:31:38human housekeepers. Correspondingly, the risks they bring are
00:31:42no longer just saying the wrong thing, but doing the wrong
00:31:45thing. What is particularly scary is when the housekeeper starts
00:31:49acting on its own, and the owner doesn't even know what
00:31:53it has done. This is the most difficult part of governance.
00:31:57How can our supervision keep up? The danger is not
00:32:01in the technology, the danger is in the lack
00:32:04of accountable technology, in having capabilities
00:32:08but no supervision, in having decisions
00:32:12but no transparency. This was the warning issued by
00:32:16UN Secretary-General Guterres on Tuesday at the opening
00:32:19of the General Debate of the UN General Assembly regarding the
00:32:23current development of artificial intelligence. The multiple AI border-crossing
00:32:27incidents that have occurred this year clearly expose
00:32:30significant loopholes in the regulatory mechanisms. Currently, regulation
00:32:34is still quite inadequate. If such experiments
00:32:38were conducted by locking an intelligent agent in a sandbox
00:32:42to observe its behavior, and if its safeguards were
00:32:45sufficient, and if the external regulatory environment was
00:32:49solid enough, then I believe the vast
00:32:53majority of such situations could be avoided.
00:32:56From the self-discipline of enterprises to the actual
00:33:00implementation platforms of government regulatory policies in the countries
00:33:03where these incidents occurred, there are some deficiencies.
00:33:07Therefore, this gives artificial intelligence systems such
00:33:10an opportunity. The AI cross-border incident behind
00:33:14it is that regulation has not kept pace with the speed of technological development. As artificial intelligence gradually acquires autonomous decision-making, tool invocation, and even
00:33:34Between governments and enterprises, there should be such
00:33:38neutral institutions, to help enterprises find
00:33:42what enterprises themselves do not have,
00:34:13in a more balanced perspective and manner.
00:34:28Since its first release in 2024, this governance
00:34:32framework has been updated for the third consecutive year.
00:34:35The new framework pays more attention to new risks
00:34:39arising from artificial intelligence, setting up a special section
00:34:43on autonomous cyber-attack threats,
00:35:12and also includes specialized risk management for intelligent agents.
00:35:47We are why do we attach importance to the process? Because in the research and development of artificial intelligence, especially in the research and development of intelligent
00:35:51agents and their use, from design, research and development,
00:35:54use, deployment,
00:35:58until the intelligent agent or artificial
00:36:01intelligence service is eliminated. At any stage of our
00:36:05entire process, different forms of risks and
00:36:09challenges may arise. That's why we emphasize
00:36:13the full life cycle and the full process. This is its
00:36:17importance. Every intelligent agent should have its
00:36:21own identity. In this way, we can prevent
00:36:25potential risks of intelligent agents in advance. On the
00:37:02other hand, it is traceability. In the process
00:37:06of solving problems, As the first institutional
00:37:10framework to directly address the risks of intelligent agents,
00:37:13Framework 3.0 has only taken the first step of
00:37:17a long march. As intelligent agents evolve from single
00:37:21intelligent agents to multiple intelligent agents, how to
00:37:25keep AI under effective supervision and human oversight
00:38:30has also become a problem faced by all countries.
00:38:34If a national government wants to control or govern
00:38:38pre-artificial intelligence, it may need to. This means
00:38:42that security in the AI field is increasingly like cybersecurity
00:38:46and nuclear security. One company's problem can
00:38:50affect a string of companies. A loophole in one country's
00:38:54regulation can also affect other countries. In the
00:38:57face of AI risks, no country can stand
00:39:01alone. No one can solve it by closing
00:39:05their doors. UN Secretary-General Guterres even equated uncontrolled
00:39:09artificial intelligence with the climate crisis, calling it
00:39:13an existential threat facing the world. Therefore, international
00:39:16cooperation is more necessary than ever before. But
00:39:20the problem is, everyone
00:39:21understands the reasoning,
00:39:24but actions may not be consistent. How should countries work together to respond?
00:39:47What worries him even more is that many people still only think about raising the tiger, but rarely think about
00:39:51how much hidden danger it will bring. In international
00:39:55relations, there is such a security dilemma.
00:40:12So, if everyone accepts this competitive illusion and thinks they want to win the competition, but the attention to security may only be a small part of the overall investment, I remember an institution estimating that it might not even reach 10% of the overall investment. If that's the case, then naturally there will be very few experts researching security
00:40:16measures, and your resource investment will also be limited.
00:40:19Then it is conceivable that this kind of security and
00:40:23development will not be balanced, right? It will
00:40:31only lead to the technology model becoming more and more
00:40:35powerful, but the security measures actually fall further
00:40:39and further behind. This is the so-called security dilemma.
00:40:42In the face of the security dilemma brought by
00:40:46artificial intelligence, international cooperation seems more urgent.
00:40:49This week, Germany and
00:40:5020 other countries and the
00:40:57European Union issued a joint statement, calling for international cooperation to ensure that AI remains under human control. It's too easy for it to cross borders.
00:42:10To promote this issue of artificial
00:42:13intelligence at a practical level is an important
00:42:17attempt for exchanges between the two countries. In the
00:42:21issue of economic and trade, risk
00:42:25prevention is a traditional issue. Therefore, it
00:42:29can combine the risk prevention of artificial
00:42:33intelligence with the development of economy and cooperation
00:42:36in terms of risk prevention in depth. So,
00:42:40I think this is also a very important
00:42:44consideration. We can also see that on the
00:42:47issue of artificial intelligence risk prevention,
00:42:51we should go beyond the economic and trade
00:42:55issue and establish a more systematic platform for Sino-US
00:42:59artificial intelligence exchanges. China and the United
00:43:03States should play a leading role in establishing such
00:43:07an international order and mechanism. On Tuesday,
00:44:10UN Secretary-General Guterres stated in the general debate
00:44:13of the UN General Assembly that we should
00:44:17strive to establish a multilateral framework for artificial intelligence
00:44:21risk management, supported by credible and independent
00:44:25supervision. World artificial. constantly exceeds our imagination.
00:44:29It opens our eyes, but also makes us
00:44:33increasingly uneasy. Some people say that innovation can
00:44:36break all boundaries. This is only half true.
00:45:34The development of artificial intelligence must