2026年人工智能的下一步是什么

欢迎收听《麻省理工科技评论》有声版。

Welcome to MIT Technology Review Narrated.

我是马特·霍纳恩。

My name is Matt Honan.

我是我们的主编。

I'm our editor in chief.

每周，我们将为您带来一篇来自科学与技术前沿的引人入胜的深度报道，涵盖人工智能、生物技术、气候、能源、机器人技术等主题。

Every week, we'll bring you a fascinating, new, in-depth story from the leading edge of science and technology, covering topics like AI, biotech, climate, energy, robotics, and more.

这是本周的故事。

Here's this week's story.

希望您喜欢。

I hope you enjoy it.

由诺亚朗读。

Narrated by Noah.

您可以在Noah应用或newsoveraudio.com上收听来自全球最大出版商的更多优质文章。

Listen to more of the best articles from the world's biggest publishers on the Noah app or at newsoveraudio.com.

威尔·道格拉斯·希文、里安娜·威廉姆斯、詹姆斯·奥唐奈尔、陈赛伟和迈克尔·金·赖特。

Will Douglas Heaven, Rhianna Williams, James O'Donnell, Saiwei Chen, and Michael Kim Wright.

2026年人工智能的下一步会是什么？

What's next for AI in 2026?

在这样一个不断变化的行业中，大胆预测未来似乎显得鲁莽。

In an industry in constant flux, sticking your neck out to predict what's coming next may seem reckless.

人工智能泡沫？

AI bubble?

什么？

What?

人工智能泡沫。

AI bubble.

但过去几年里，我们一直这么做，现在我们又将再次这样做。

But for the last few years, we've done just that, and we're doing it again.

上次我们预测得怎么样？

How did we do last time?

我们挑选了2025年值得关注的五大热门AI趋势，其中包括我们所说的生成式虚拟游乐场，即世界模型。

We picked five hot AI trends to look out for in 2025, including what we called generative virtual playgrounds, aka world models.

从谷歌DeepMind的Genie 3到WorldLabs的Marvel，能够实时生成逼真虚拟环境的技术正变得越来越好。

Check from Google DeepMind's Genie three to WorldLabs Marvel, tech that can generate realistic virtual environments on the fly keeps getting better and better.

所谓的推理模型，没错，还需要多说吗？

So called reasoning models, check, need we say more?

推理模型已迅速成为顶尖问题解决的新范式。

Reasoning models have fast become the new paradigm for best in class problem solving.

AI在科学领域的蓬勃发展。

A boom in AI for science.

没错。

Check.

OpenAI现在也效仿谷歌DeepMind，组建了一个专门团队专注于这一领域。

OpenAI is now following Google DeepMind by setting up a dedicated team to focus on just that.

与国家安全关系更密切的AI公司。

AI companies that are cozier with national security.

检查。

Check.

OpenAI 改变了其技术用于战争的立场，与国防科技初创公司 Andoril 达成协议，帮助其击落战场无人机。

OpenAI reversed position on the use of its technology for warfare to sign a deal with the defense tech startup, Andoril, to help it take down battlefield drones.

对 NVIDIA 的正当竞争。

And legitimate competition for NVIDIA.

检查，有点吧。

Check, kind of.

中国正在全力开发先进 AI 芯片，但至少目前 NVIDIA 的主导地位仍难以撼动。

China is going all in on developing advanced AI chips, but NVIDIA's dominance still looks unassailable, for now at least.

那么 2026 年会有什么新动向？

So what's coming in 2026?

以下是我们在未来十二个月的重大预测。

Here are our big bets for the next twelve months.

第一，会有更多硅谷产品基于中国大语言模型构建。

One, more Silicon Valley products will be built on Chinese LLMs.

过去一年，中国开源模型取得了重大进展。

The last year shaped up as a big one for Chinese open source models.

2025年1月，深度求索发布了其开源推理模型R1，凭借有限的资源，这家中国小型公司震惊了世界。

In January 2025, DeepSeek released R1, its open source reasoning model, and shocked the world with what a relatively small firm in China could do with limited resources.

到年底，'DeepSeek时刻'一词已成为人工智能创业者、观察者和开发者口中常提及的短语，成为一种理想化的标杆。

By the end of the year, DeepSeek moment had become a phrase frequently tossed around by AI entrepreneurs, observers, and builders, an aspirational benchmark of sorts.

这是许多人第一次意识到，无需通过OpenAI、Anthropic或Google，也能体验到顶级AI性能。

It was the first time many people realized they could get a taste of top tier AI performance without going through OpenAI, Anthropic, or Google.

像R1这样的开源模型允许任何人下载并在自己的硬件上运行。

Open weight models like R1 allow anyone to download a model and run it on their own hardware.

它们也更具可定制性，团队可以通过蒸馏和剪枝等技术对模型进行调整。

They are also more customizable, letting teams tweak models through techniques like distillation and pruning.

这与美国大型公司发布的闭源模型形成鲜明对比，后者的核心能力仍属专有，且访问成本往往高昂。

This stands in stark contrast to the closed models released by major American firms, where core capabilities remain proprietary and access is often expensive.

因此，中国模型已成为一个轻松的选择。

As a result, Chinese models have become an easy choice.

CNBC和彭博社的报道表明，美国初创企业越来越认识到并接受了中国模型所提供的优势。

Reports by CNBC and Bloomberg suggest that startups in The US have increasingly recognized and embraced what they can offer.

其中一类广受欢迎的模型是Qwen，由阿里巴巴创建，阿里巴巴是中国最大的电商平台淘宝的母公司。

One popular group of models is Quen, created by Alibaba, the company behind China's largest e commerce platform Taobao.

仅Qwen 2.5-1.5b-instruct这一版本就已获得885万次下载，成为使用最广泛的预训练大语言模型之一。

Quen 2.5-1.5b instruct alone has 8,850,000 downloads, making it one of the most widely used pre trained LLMs.

Qwen系列涵盖多种模型规模，并针对数学、编程、视觉和指令遵循等任务进行了专门优化。

The Quinn family spans a wide range of model sizes alongside specialized versions tuned for math, coding, vision, and instruction following.

这种多样性使其成为开源领域的巨头。

A breadth that has helped it become an open source powerhouse.

其他曾对开源持观望态度的中国AI公司，如今也开始效仿DeepSeek的策略。

Other Chinese AI firms that were previously unsure about committing to open source are following DeepSeq's playbook.

突出代表包括智谱的GLM和月之暗面的Kimi。

Standouts include Jipu's GLM and Moonshot's Kimi.

这种竞争也促使美国公司至少在一定程度上开放了其技术。

The competition has also pushed American firms to open up, at least in part.

2025年8月，OpenAI发布了其首个开源模型。

In August 2025, OpenAI released its first open source model.

11月，位于西雅图的非营利组织艾伦人工智能研究所发布了其最新的开源模型OMO 3。

In November, the Allen Institute for AI, a Seattle based nonprofit released its latest open source model, OMO three.

尽管美中对抗加剧，中国人工智能公司对开源的近乎一致支持，为它们赢得了全球人工智能界的善意和长期信任优势。

Even amid growing US China antagonism, Chinese AI firms near unanimous embrace of open source has earned them goodwill in the global AI community and a long term trust advantage.

2026年，预计将有更多硅谷应用程序在中文开源模型之上悄然运行，并关注中国发布与西方前沿之间的差距从数月缩短至数周，有时甚至更短。

In 2026, expect more Silicon Valley apps to quietly ship on top of Chinese open models and look for the lag between Chinese releases in the Western frontier to keep shrinking from months to weeks and sometimes less.

第二，美国将迎来又一年的监管拉锯战。

Two, The US will face another year of regulatory tug of war.

关于人工智能监管的斗争正走向对决。

The battle over regulating artificial intelligence is heading for a showdown.

2025年12月11日，总统唐纳德·特朗普签署了一项行政命令，旨在削弱各州的AI法规，此举意在束缚各州对快速发展的行业进行监管的能力。

On 12/11/2025, president Donald Trump signed an executive order aiming to neuter state AI laws, a move meant to handcuff states from keeping the growing industry in check.

2026年，预计将出现更多政治博弈。

In 2026, expect more political warfare.

白宫与各州将就谁有权监管这一蓬勃发展的技术展开争斗，而人工智能公司则展开激烈的游说活动，试图扼杀监管，其论点是：各州法律的碎片化将扼杀创新，并使美国在与中国的AI军备竞赛中处于劣势。

The White House and states will spar over who gets to govern the booming technology, while AI companies wage a fierce lobbying campaign to crush regulations, armed with the narrative that a patchwork of state laws will smother innovation and hobble The US in the AI arms race against China.

根据特朗普的行政命令，各州可能担心因与他倡导的宽松监管理念相冲突而面临诉讼或联邦资金被切断。

Under Trump's executive order, states may fear being sued or starved federal funding if they clash with his vision for light touch regulation.

像加利福尼亚州这样的大型民主州刚刚通过了美国首部前沿人工智能法，要求公司公布其AI模型的安全性测试结果，它们将把这场斗争带入法庭，主张只有国会才有权推翻州法律。

Big Democratic states like California, which just enacted the nation's first frontier AI law requiring companies to publish safety testing for their AI models, will take the fight to court, arguing that only Congress can override state laws.

但那些负担不起失去联邦资金或害怕触怒特朗普的州可能会退缩。

But states that can't afford to lose federal funding or fear getting in Trump's crosshairs might fold.

然而，我们仍会看到更多州在热点问题上立法，尤其是在特朗普的命令为州政府立法亮起绿灯的领域。

Still expect to see more state lawmaking on hot button issues, especially where Trump's order gives states a green light to legislate.

由于聊天机器人被指控引发青少年自杀，而数据中心消耗的能量越来越多，各州将面临日益增长的公众压力，被迫推动建立监管框架。

With chatbots accused of triggering teen suicides and data centers sucking up more and more energy, states will face mounting public pressure to push for guardrails.

作为州法律的替代方案，特朗普承诺将与国会合作制定联邦人工智能法。

In place of state laws, Trump promises to work with Congress to establish a federal AI law.

别指望它能实现。

Don't count on it.

国会曾在2025年两次未能通过暂停各州立法的提案，我们并不指望它能在2026年推出自己的法案。

Congress failed to pass a moratorium on state legislation twice in 2025, and we aren't holding out hope that it will deliver its own bill in 2026.

像OpenAI和Meta这样的AI公司将继续动用强大的超级政治行动委员会，支持支持其议程的政治候选人，并打击那些阻碍他们的人。

AI companies like OpenAI and Meta will continue to deploy powerful super PACs to support political candidates who back their agenda, and target those who stand in their way.

另一方面，支持AI监管的超级政治行动委员会也将建立自己的资金库进行对抗。

On the other side, super PACs supporting AI regulation will build their own war chests to counter.

关注他们在2027年中期选举中的激烈较量。

Watch them duke it out at twenty twenty seven's midterm elections.

AI越发展，人们就越会努力引导其发展方向，2026年将继续是监管拉锯战的一年，且看不到尽头。

The further AI advances, the more people will fight to steer its course, and 2026 will be another year of regulatory tug of war, with no end in sight.

三、聊天机器人将改变我们的购物方式。

Three, chatbots will change the way we shop.

想象这样一个世界：你拥有一个随时待命的私人购物助手，这位专家能立即为你推荐适合最难挑选的亲友的礼物，或上网搜索出符合你紧张预算的最佳书架清单。

Imagine a world in which you have a personal shopper at your disposal 20 an expert who can instantly recommend a gift for even the trickiest to buy for a friend or relative, or trawl the web to draw up a list of the best bookcases available within your tight budget.

更棒的是，它们还能分析厨房电器的优缺点，将其与看似相同的竞品进行对比，帮你找到最划算的交易。

Better yet, they can analyze a kitchen appliance's strengths and weaknesses, compare it with its seemingly identical competition, and find you the best deal.

一旦你对他们的建议感到满意，他们还会帮你处理购买和配送的细节。

Then once you're happy with their suggestion, they'll take care of the purchasing and delivery details too.

但这个极其博学的购物助手根本不是什么精明的人类。

But this ultra knowledgeable shopper isn't a clued up human at all.

它是一个聊天机器人。

It's a chatbot.

这也不是什么遥远的预测。

This is no distant prediction either.

Salesforce最近表示，预计今年假日季AI将推动2630亿美元的在线消费。

Salesforce recently said it anticipates that AI will drive $263,000,000,000 in online purchases this holiday season.

这约占所有订单的21%。

That's some 21 percent of all orders.

专家们预测，未来几年AI增强的购物将成为更大的生意。

And experts are betting on AI enhanced shopping becoming even bigger business within the next few years.

根据咨询公司麦肯锡的研究，到2030年，代理式商业的年交易额将达到3万亿至5万亿美元。

By 2030, between 3 and $5,000,000,000,000 annually will be made from Agentic Commerce, according to research from the consulting firm McKinsey.

毫不意外，人工智能公司已经大力投资，力求让通过其平台进行购买的过程尽可能顺畅。

Unsurprisingly, AI companies are already heavily invested in making purchasing through their platforms as frictionless as possible.

谷歌的Gemini应用现在可以访问公司强大的产品和卖家购物图数据集，甚至能使用其代理技术代表你致电商店。

Google's Gemini app can now tap into the company's powerful shopping graph data set of products and sellers, and can even use its agentic technology to call stores on your behalf.

与此同时，早在2025年11月，OpenAI就宣布了一项ChatGPT购物功能，能够快速生成购物指南，并已与沃尔玛、塔吉特和Etsy达成合作，允许购物者直接在聊天机器人互动中购买商品。

Meanwhile, back in November 2025, OpenAI announced a chat GPT shopping feature capable of rapidly compiling buyers guides, and the company has struck deals with Walmart, Target, and Etsy to allow shoppers to buy products directly within chatbot interactions.

随着消费者与人工智能聊天的时间持续增加，而来自搜索引擎和社交媒体的网页流量持续下降，预计未来一年内将出现更多此类合作。

Expect plenty more of these kinds of deals to be struck within the next year, as consumer time spent chatting with AI keeps on rising, and web traffic from search engines and social media continues to plummet.

第四，大型语言模型将做出一项重要新发现。

Four, an LLM will make an important new discovery.

我一开始就先打个预防针，但众所周知，大型语言模型会生成大量无意义的内容。

I'm going to hedge here right out of the gate, but it's no secret that large language models spit out a lot of nonsense.

除非有猴子和打字机那样的运气，否则大型语言模型自身无法独立发现任何新事物，但它们确实仍有潜力拓展人类知识的边界。

Unless it's with monkeys and typewriters luck, LLMs won't discover anything by themselves, but LLMs do still have the potential to extend the bounds of human knowledge.

2025年5月，谷歌DeepMind公布了AlphaEvolve系统，该系统利用其Gemini大型语言模型，提出了用于解决未解问题的新算法，让我们一窥这种可能性如何实现。

We got a glimpse of how this could work in May 2025, when Google DeepMind revealed AlphaEvolve, a system that used the firm's Gemini LLM to come up with new algorithms for solving unsolved problems.

这一突破在于将Gemini与一种进化算法相结合，该算法会检查其建议、挑选最优方案，并将它们反馈给大语言模型以进一步优化。

The breakthrough was to combine Gemini with an evolutionary algorithm that checked its suggestions, picked the best ones, and fed them back into the LLM to make them even better.

谷歌DeepMind利用AlphaEvolve找到了更高效管理数据中心和谷歌TPU芯片功耗的方法。

Google DeepMind used AlphaEvolve to come up with more efficient ways to manage power consumption by data centers and Google's TPU chips.

这些发现意义重大，但尚未带来颠覆性改变。

Those discoveries are significant, but not game changing yet.

谷歌DeepMind的研究人员正在推动这一方法，以探索其潜力极限，而其他人也迅速跟进。

Researchers at Google DeepMind are now pushing their approach to see how far it will go, And others have been quick to follow their lead.

AlphaEvolve发布一周后，新加坡的AI工程师阿桑基·阿什玛分享了OpenEvolve，这是谷歌DeepMind工具的开源版本。

A week after AlphaEvolve came out, Asanki Asharma, an AI engineer in Singapore, shared OpenEvolve, an open source version of Google DeepMind's tool.

2025年9月，日本公司Sakana AI发布了名为SyncAevolve的软件版本。

In September 2025, the Japanese firm Sakana AI released a version of the software called SyncAevolve.

11月，一支由美国和中国研究人员组成的团队公布了Alpha Research，他们声称该工具改进了AlphaEvolve中一项已优于人类的数学解决方案。

And in November, a team of US and Chinese researchers revealed Alpha research, which they claim improves on one of AlphaEvolve's already better than human math solutions.

还有其他替代方法。

There are alternative approaches too.

例如，科罗拉多大学丹佛分校的研究人员正试图通过调整所谓推理模型的工作方式，使大语言模型更具创造力。

For example, researchers at the University of Colorado Denver are trying to make LLMs more inventive by tweaking the way so called reasoning models work.

他们借鉴了认知科学家对人类创造性思维的了解，推动推理模型提出更跳出框架、而非通常那种保守的建议。

They have drawn on what cognitive scientists know about creative thinking in humans to push reasoning models toward solutions that are more outside the box than their typical safe bet suggestions.

数百家公司正在投入数十亿美元，寻找让人工智能解决未解数学难题、加快计算机速度以及开发新药物和新材料的方法。

Hundreds of companies are spending billions of dollars looking for ways to get AI to crack unsolved math problems, speed up computers, and come up with new drugs and materials.

既然AlphaEvolve已经展示了大语言模型的潜力，预计这一领域的活动将迅速增加。

Now that AlphaEvolve has shown what's possible with LLMs, expect activity on this front to ramp up fast.

五、法律纠纷加剧。

Five, legal fights heat up.

一段时间以来，针对人工智能公司的诉讼一直相当可预测。

For a while, lawsuits against AI companies were pretty predictable.

作者或音乐家等权利持有人会起诉那些使用其作品训练AI模型的公司，而法院通常倾向于支持科技巨头。

Rights holders like authors or musicians would sue companies that trained AI models on their work, and the courts generally found in favor of the tech giants.

人工智能即将面临的法律斗争将远比以往更加复杂。

AI's upcoming legal battles will be far messier.

这些争端集中在棘手且未解决的问题上。

The fights center on thorny, unresolved questions.

AI公司是否应对聊天机器人鼓励人们采取的行为负责，例如当它们帮助青少年策划自杀时？

Can AI companies be held liable for what their chatbots encourage people to do, as when they help teens plan suicides?

如果聊天机器人散布了关于你的明显虚假信息，其创建者是否可以被起诉诽谤？

If a chatbot spreads patently false information about you, can its creator be sued for defamation?

如果公司败诉，保险公司是否会拒绝将AI公司作为客户？

If companies lose these cases, will insurers shun AI companies as clients?

到2026年，我们将开始看到这些问题的部分答案，部分原因是某些著名案件将进入审判阶段。

In 2026, we'll start to see the answers to these questions, in part, because some notable cases will go to trial.

一名自杀身亡青少年的家人将于十一月将OpenAI告上法庭。

The family of a teen who died by suicide will bring OpenAI to court in November.

与此同时，特朗普总统于2025年12月发布的行政命令将进一步复杂化法律环境。

At the same time, the legal landscape will be further complicated by President Trump's executive order from December 2025.

无论如何，我们将看到各种方向上令人眼花缭乱的诉讼，更不用说一些法官在海量案件中甚至转向求助AI了。

No matter what, we'll see a dizzying array of lawsuits in all directions, not to mention some judges even turning to AI amid the deluge.

您正在收听《麻省理工科技评论》，本文由威尔·道格拉斯·海文、里安农·威廉姆斯、詹姆斯·奥唐奈尔、蔡伟臣和米歇尔·金撰写。

You were listening to MIT Technology Review, where Will Douglas Heaven, Riannon Williams, James O'Donnell, Tsai Wei Chen, and Michelle Kim write.

2026年人工智能的下一步会是什么？

What's next for AI in 2026?

本文发布于2026年1月5日，由萨姆·肖尔为诺亚朗读。

This article was published on the 01/05/2026, and was read by Sam Scholl for Noah.