- 23分钟前
八点最热报 | 美国财长贝森特亲口承认，中国的AI模型水平已经追到美国的八至九成。他提议，美中之间建立“AI安全通报机制”，以确保如果哪天出了事也会有人接电话，而且他很有信心中国会点头。从过去试图封杀并将中国挡在技术壁垒之外，到如今承认中国在AI领域的成就，并提出建立AI安全通报机制，中美这场算力杀局，已经出现新的变化。（主播: 梁宝仪）
类别
🗞
新闻消息文字稿
00:00Before watching the video, remember to check out trending websites for more content!
00:03US Treasury Secretary Bessant himself admitted
00:06China's AI model level has caught up with 80% to 90% of that of the United States.
00:11The proposal to establish an AI security notification mechanism between the US and China
00:15This is to ensure that if something happens someday, someone will answer the phone.
00:19Moreover, he was very confident that China would agree.
00:22From past attempts to block and keep the Chinese Communist Party outside all barriers
00:27Now I've accepted the facts.
00:29Acknowledging China's achievements in the field of AI
00:31It was proposed to establish an AI safety notification mechanism.
00:35The computing power battle between China and the US has taken a new turn.
00:39But this statement also indirectly acknowledges that eight to nine out of ten people are guilty.
00:44Chinese AI has made it to the final battle.
00:47This has completely exposed the United States' long-standing confidence in the field of AI.
00:52The script in the United States in recent years is written very clearly.
00:54To control high-end chips
00:56To lock down computing power and ensure China always catches up.
00:59However, the US now publicly acknowledges that China's AI breakthrough is already 90% complete.
01:05combat power
01:05Has the United States been forced to sound a security alarm?
01:09From absolute blackout to a hotline for requesting audiences
01:13What is it that the US is truly afraid of about China?
01:17A myth of America built on absolute dominance
01:20That was the first time a hole was poked by one of his own officials.
01:23So why is it that they're demanding a talk now of all times?
01:26Please call the hotline.
01:29The reason is actually not hard to understand.
01:31For example
01:32The two race cars were speeding side by side on the same mountain road.
01:35No one was willing to let go of the accelerator.
01:37Their braking systems stopped communicating with each other.
01:41The most dangerous thing is not man
01:42It was almost time for an accident to happen.
01:45There isn't even a channel to call for a halt.
01:48That's the brake cable that Bessenter wants to install.
01:52However, when today's AI models go out of control
01:55Cyberattack misjudged by automated military system
01:57Is this brake cable really effective at stopping the car?
02:01Tearing away the disguise of security
02:04The US Treasury Secretary proactively discussed...
02:06China's AI capabilities have reached an astonishing level.
02:09What kind of rule-based coercion is hidden behind this?
02:16U.S. Treasury Secretary Bessenter
02:18With Chinese Vice Premier He Lifeng
02:20Earlier high-level meetings were held
02:21A key focus is on promoting the establishment of a US-China AI security communication channel.
02:25After the meeting between the two sides
02:26America’s anxiety is obvious
02:28When the reins of technology may slip out of your hands
02:30Washington urgently needs a hotline directly connecting to Beijing.
02:33Because for Washington
02:35Our leading position in AI must not be relinquished.
03:01The narrow lead is making Washington walk on thin ice.
03:04Facing the pursuers, China
03:06The United States is not only concerned with who is in the lead.
03:09More attention is paid to what happens when massive computing power is no longer subject to red line constraints.
03:13What kind of chain reaction will the world face?
03:16In fact, the United States' concerns about security are not unfounded.
03:19From Washington's perspective
03:21Even if China's AI models only reach 90% of the level of the United States
03:24If protection is lacking
03:26It's like a vehicle loaded with refined fuel
03:28However, the racing car lost its braking system
03:30Cyberattacks could occur at any time.
03:32Giant winds may be triggered in areas such as biological threats.
04:00If the technological competition between major powers in the past was a race to see who could win the most rice?
04:04race
04:05So today's AI competition has turned into a two-horse race on the edge of a cliff.
04:10Human skating
04:10Both sides not only need to compete on the difficulty of the moves
04:13We must always be vigilant to prevent the other party, or even ourselves, from slipping off the edge of the cliff.
04:17However, at a time when strategic mutual trust among major powers has reached its lowest point...
04:21A so-called AI security hotline
04:24Can it truly become a barrier against ultimate risk?
04:27The US-China AI race is still ongoing.
04:29The rules of the game may be rewritten.
04:31Can the two countries find a balance between security and development?
04:34Find that delicate balance
04:36The answer may lie in this trans-Pacific game of computing power and security.
04:40Among
04:44In the nuclear weapons era
04:45Communication itself is a safety device
04:47If today's AI encounters problems...
04:50Are the two sides able to communicate immediately?
04:52This logic
04:54Many scholars who study great power competition have long said...
04:57Harvard scholar Allison's Thucydides Trap
05:01The core issue is the choice between the decisive and the conservative.
05:04They are often not dragged into a deliberate war.
05:08Instead, it was pushed down by misjudgment and accidents.
05:11When the myth of technology is shattered
05:13Admitting that China's AI has caught up to 90%
05:15The US-China AI competition has reached a precipice.
05:18So, is this hotline really just a safety valve?
05:22Crisis in the Eye of the Storm
05:24What are the motives behind this US suggestion?
05:27When a country's leading advantage
05:29The difference between the crushing and the compression has been reduced to only 10%.
05:32The most rational approach is not to continue the head-on confrontation.
05:35Instead, it incorporates competition into the rules.
05:38Keep the power to draft the rules in your own hands.
05:42Today we'll discuss the notification.
05:44We'll talk about standards tomorrow.
05:46We'll discuss whose model is the safest the day after tomorrow.
05:49To put it bluntly, who defines security?
05:51So who held the reins?
05:53The United States used to try to use a wall to stop China's catch-up.
05:57As a result, the higher the wall...
05:59China has developed the ability to bypass internet restrictions and catch up.
06:03Now the walls can't stop it.
06:05Turning around, I wanted to repair a guardrail.
06:07But true safety doesn't depend on who calls for a halt first.
06:10Instead, it relies on neither side daring to
06:12There's no need to floor the gas pedal.
06:14However, the hotline proposed by the United States
06:17Is it admitting defeat or a social issue?
06:19What the US needs is a single, decisive step that can put a stop to its adversaries.
06:22Is it possible to control the brakes on both sides in one step?
06:25The question might not be on the other end of the phone.
06:28And whoever is willing to put down their arrogance first