瑞典AI改革介绍
主持人: 大家好。今天我请到的是 Sana 的创始人兼CEO Joel Helmar。Sana是一家非常重要的瑞典AI公司,专注于知识管理。Joel Helmar是备受瞩目的瑞典AI改革的推动者。这项改革将把AI模型带给230万瑞典人。
Original English
Interviewer: Hi everybody. Today I'm joined by Joel Helmar who is the founder and CEO of Sana which is an a really important Swedish AI company. They specialize in knowledge management and Joel is behind the very important new Swedish AI reform. Now this reform will bring AI models to 2.3 million Swedes. Wow.
Joel Helmar: 这是什么瑞典AI改革?瑞典AI改革的历史可以追溯到90年代,当时瑞典推出了PC改革。PC改革的重点是为每个家庭提供一台个人电脑。通过巧妙的税务设置,我们成功让一百万户家庭拥有了电脑,许多人认为瑞典的科技奇迹,如果可以这么说的话,就源于这次PC改革。所以我们开始思考,AI的对等措施会是什么样子?我们能为AI做些什么,使其在瑞典普及化?因此,我们启动了瑞典AI改革。我们免费为230万瑞典人提供尖端模型。这包括所有大学、所有学校、所有公职人员都能免费使用所有尖端模型。
Original English
Joel Helmar: What is the Swedish AI reform? The history of the Swedish AI reform goes back to the 90s where Sweden introduced the PC reform. The PC reform was was focused on providing a a personal computer in every household. Um through um you know clever tax uh setups, we managed to get one million households with pieces and a lot of folks credit you know the Swedish tech wonder if you like coming from the Swedish PC reform. So what we started thinking about is what does the equivalent for AI look like? What could we do for AI to make it democratic across Sweden? And so we set up uh the Swedish AI reform. We're providing frontier models to 2.3 million Swedes for free. So that's all universities, all schools, all public servants that gets free access to all of the frontier models.
改革的资金来源
主持人: 谁来支付这笔费用?
Original English
Interviewer: Who pays for this?
Joel Helmar: 我们决定建立一个由基金会资助的结构。通过瑞典一些最成功的商界人士的慷慨支持,我们设立了一个基金会,最终资助了这些模型的访问权限,使我们能够超快速地推进。在短短六周内,我们完成了基金会的设立、融资,并开始向瑞典民众分发这些模型。
Original English
Joel Helmar: So we decided to set up a structure uh where the foundation paid for it. So through the generous support of um some of the Swedish some of Sweden's most successful business folks uh we set up a foundation that ultimately funded uh the the access to these models that enabled us to move super quickly. So uh during the time of six weeks we managed to set up the foundation get it funded and and then start distributing these models to to to the Swedish population.
资金规模与公私协作
主持人: 如果要给个数字,大概是多少钱?
Original English
Interviewer: And how much money are we talking here if you have to put a number?
Joel Helmar: 对于像瑞典这样的国家来说,分发成本大约在数百万美元。当然,这还取决于服务的具体使用量。
Original English
Joel Helmar: I think for a nation like like like Sweden um it's in the sort of millions of of dollars to to distribute it but then of course it depends on the the usage of the services.
主持人: 所以是2000万到3000万美元?
Original English
Interviewer: So 20 30 million Swedish.
Joel Helmar: 是的,随着规模的扩大,可能会有这个范围,但这是一个大概的数字。
Original English
Joel Helmar: Yeah it it can can range uh as you scale it up but that would be a ballpark.
主持人: 你们如何划分私营部门和公共部门的工作?
Original English
Interviewer: How do you split the work between the private sector and the public sector here?
Joel Helmar: 私营部门负责资助,也负责构建工具。我们与公共部门合作进行分发。我们希望关注的是,能否获得推动这些倡议所需的资金速度,能否获得开发速度和开发质量。如果这些工具不是消费级工具,人们就不会采纳它们,它们需要处于AI的最前沿。但如果我们能将其与公共部门结合起来,公共部门自然是我们社会的支柱,拥有所有的分发渠道和所有的信任。那将是最有效的地方。所以,我们负责融资。我不想等待政府批准来资助此事。我不想等待漫长的政府开发这些工具的流程。我认为这是我们可以在私营部门做得很好的地方。然而,这需要大量的信任。这需要大量的分发能力,而我认为公共部门在这方面做得非常好。
Original English
Joel Helmar: So the private sector is is funding it. The private sector is also building the tools and we're collaborating with the public sector to distribute it and so um what what we want to focus on is can we get the speed of of funding that we can get for these initiatives can we get the speed of the development the quality of the development uh if these tools are not consumer grade tools people are not going to adopt them they they need to be at the very frontier of of AI too but then if we can combine that with the public sector that of course powers our societies and has all of the distribution and has all of the trust. Um, that's where it can be the most effective. So, we're taking care of the funding. I don't want to wait for government approvals to fund this. I don't want to wait for a long process for governments to develop these tools. That's what I think we can do great in the in in the in the private sector. However, there's a lot of trust required for this. There's a lot of distribution required for this and that's what I think the public sector can do can do really well.
分发机制与模型多样性
主持人: 公共部门愿意分发吗?他们在瑞典分发得好吗?
Original English
Interviewer: Is the public sector willing to distribute this? Are they doing a good job distributing it in Sweden?
Joel Helmar: 我认为是这样。我们有两种方法。一部分是自下而上,一部分是自上而下。自下而上,任何人都可以访问网站进行注册,立即获得访问权限,他们可以使用AI工具来支持日常工作。还有其他方面,我们采取更自上下的方式,与不同政府机构的领导者进行沟通,然后他们会在其组织内进行分发。
Original English
Joel Helmar: Uh I I I think so. And we we have two approaches. So um it's part bottoms up, part tops down. So uh bottoms up, anyone can sign up on the website, just instantly get get access, they get access to these AI tools that can support them in their daily work. And then there's other aspects where we're doing more top down where we're talking with with the leaders of different government um institutions and then they are distributing it in in their in their organizations.
主持人: 那么,简单来说,这个工具基本上是你注册了你的工具,而这个工具背后有不同类型的模型,对吗?它不只依赖于一个大型语言模型。它就是一系列工具。我的理解正确吗?
Original English
Interviewer: So just to get it straight the tool basically you you uh you signed up to your tool and behind the tool you have different type of models, right? So it's not dependent just on one uh one uh large language model. It's just like a range of tools. Is that correct understanding?
Joel Helmar: 完全正确。你可以访问所有尖端模型。无论是你想在Anthropic、Gemini还是OpenAI的模型上运行,我们都能访问所有尖端模型。这些模型在某些情况下,每月访问费用可能高达数百美元。此外,我们还允许你将这些模型连接到你的数据或知识。我们以一种尊重你可能在公共部门所需的所有安全措施的方式构建了这个系统。
Original English
Joel Helmar: Exactly. So you get access to all of the frontier models. Whether you want to run on anthropic or or Gemini or open AI models or anything else, we have access to all of the front frontier uh frontier models. These are models that in some cases can cost you hundreds of dollars a month to get get access to. Combined with that, we also enable you to connect those models to your data or your knowledge. And we've built this system in a way that respects all of the guard rails that you for example need in public sector.
AI普及的长期影响
主持人: 考虑到这一点,我有点不太明白,瑞典这里有了所有这些工具,而挪威却没有应用这些工具,这两个国家会发生什么?
Original English
Interviewer: So given that I'm a bit thick u what's going to happen Sweden here with all the tools Norway here not applying the tools what's going to happen to the two countries
Joel Helmar: 我认为你会获得使用这些工具的复合优势,这种优势会随着时间的推移不断累积。因此,我的期望是,更有效地采用这些工具的公司将拥有更成功的企业,包括新公司的成立。这正是PC改革之后发生的情况。人们学会了编程,然后当他们学会编程时,他们创立了Spotify等公司。其次,瑞典公司将更具竞争力,因为他们知道如何使用这些工具。最终,它会影响你的利润率。你将能够提供更好、更具成本效益的产品。第三,我认为作为一个社会,我们将拥有更高质量的医疗保健、更高质量的教育和更高质量的研究。而且这一切每年都会不断累积。因此,我认为那些采纳速度较慢的国家将需要很长时间才能赶上来。
Original English
Joel Helmar: I think you get a compounding advantage of using these tools and that advantage will just compound uh over over time. So um my expectations is the companies that adopt these tools more effectively, they're going to have more successful companies, both new companies being founded. That's what happened after the PC reform. People learned how to program and then when they learned how to program, they founded Spotify and Cla and and other companies like like that. But secondly, the Swedish companies are going to be more competitive because they know how to use these tools. They can, you know, ultimately it starts eating at at your margins. are going to be able to provide a better better better product more cost- effectively. Um and then third, I think um as a society, we're going to have higher quality healthcare, higher quality education, and higher quality research. And this every single year will just keep compounding. And so I think the countries that adopt this slower will will take a very long time to to catch