Open AI的CEO和创始人山姆·阿尔特曼Sam Altman在2022年9月22日的一次采访中,Open AI的CEO Sam Altman分享了他对人工智能现在和未来的看法。

第一个预测，AI可以自己进行科学发现。AI目前的形式越来越擅长指引方向，但是如果没有人告诉它，它不能独立思考或真正做任何事情。例如，它不会自己出去解决癌症，但作为助手，它有可能帮助人类更接近这个目标。所以这里的预测是，有一天这种情况会改变，AI将能够开始自己的新科学发现，可能解决复杂的健康问题、数学悖论、清洁能源突破等等。

预测2，AI带来无限能源和无限智力。人类社会在历史上一直受到可用能源和可用智力或知识的限制。实际上，我们只能做到我们有限精力去做某事的程度，或者我们不知道如何去做。例如，直到1969年，当我们的能源输出水平和知识水平都达到一定程度时，才可能登上月球。AI将极大地影响我们现有的知识库或认知输出，它可能会影响能源行业和所有智力相关行业。想象一下，如果从现在起短短10年内，我们利用核聚变的力量，能够在全球部署清洁和几乎无限的能源，会发生什么？这会对社会产生什么深远的影响？癌症的大幅减少甚至消除呢？

预测3，在未来几年，AI将作为生产力助手在企业和商业中得到普遍应用。我认为这是最有可能在最短的时间内实现的预测。openai已经通过chatgpt证明，AI聊天机器人可以通过几种有意义的方式大大提高生产力。不久之后，我们可能会看到b2b公司开始构建专门用于各种行业的AI助手，甚至为公司提供根据自己的私人数据培训这些AI助手的能力。

预测4，下一个万亿美元的公司将建立在AI之上。也许是一家AI公司从根本上改善了我们搜索数据和与互联网互动的方式，颠覆掉了谷歌和百度这种传统搜索公司；或者是一家利用AI快速生产各种开创性创新药的生物技术公司，甚至是一家利用AI开发可扩展核聚变反应堆的能源公司。AI将推动人类前进，就像几百年前的工业革命一样。

预测5，AI大模型将成为大部分公司AI应用的基础设施。大多数公司将使用少数经过大规模训练的大模型作为其AI系统的基础。这些公司只用在这些大模型之上构建或微调它们，而不是从头开始构建自己的AI模型。那么这意味着什么呢？对于通过将大量文本输入计算机而构建的AI模型，也称为大语言模型，这个过程非常耗时、昂贵和艰苦。所以为什么要重新发明轮子或从头开始构建？

第六个预测，所与人都会去学习使用AI工具，AI会像互联网一样全面融入我们的生活。不管我们喜不喜欢，AI的进步是不可避免的，而且可能会加速。担心这可能对我们的工作保障产生的影响是没有用的。让我们这个物种如此特别的一件事是，我们是适应进化的大师。这次也不例外。因此，尽你所能学习利用这些新AI工具来为自己谋利。

预测7，融合文字图片语音视频的多模态AI模型将在未来几年到来。今天，AI越来越擅长做单一的任务。某些AI可以从文本提示中创建图像，某些AI可以创建对问题的对话回答。但很少可以同时完成两项以上的任务。所谓的能力，是通过单个AI模型或接口来做多件事。理想的状况是：如果我要求AI创建一份关于经济状况的文本报告，包括圆周率和条形图以及支持数据，所有这些都通过一个界面，还可以让它生成视频内容。

第八个也是最后一个预测是，AI将推动组织回到线下办公室。未来十年最具创新性和最有价值的公司将是那些重新融入办公室工作模式的公司，而非远程办公。

英文原版：

Hey, YouTube. In this video, I'm going to take you through several predictions on the future of AI by sam Altman, the CEO and co founder of open AI. Now, if you're not familiar with openai, it's an artificial intelligence research and development company founded in 2015 by Elon Musk, sam Altman and several others. So with the recent popularity of chatchie, BT and dolly, sam's been doing the rounds and talking a little bit about how he sees the future of AI. So I wanted to make this video to decide still some of those predictions, as well as to add some of my own thoughts. As always, if you liked this video, please consider liking and subscribing below as it really helps me grow the channel. All right, let's get started.

Prediction number one, AI in its current form is getting really good at taking directions, but it can't think for itself or really do much of anything without a human telling it to. For example, it won't go out and solve cancer by itself, but it will potentially help humans get closer to that goal as an assistant. So the prediction here is that one day this will change and AI will be able to begin making new scientific discoveries of its own, possibly solving complex health problems, mathematical paradoxes, clean energy breakthroughs, etc.

Prediction number 2, this one's really fun and kind of meta. So human societies have historically always been limited by 2, available energy and available intelligence or knowledge. Effectively, we've only been able to do things up to the point where we don't have the energy to do something, and or we don't know how to do it. Going to the moon, for example, was not possible until 1969, when both our levels of energy and knowledge had reached a certain point. AI is about to drastically impact our pool of available knowledge or cognitive output, and it's likely to impact energy as well as a thought experiment. Imagine what would happen if in just 10 short years from now, we harness the power of nuclear fusion and are able to deploy clean and nearly limitless energy across the globe. What profound impacts would that have on society? What about the drastic reduction or even elimination of cancer?

Interesting to think about, to say the least. Prediction number 3, over the next few years, AI will become prevalent in the business sector as a productivity assistant. Now, I've talked about this one in my other videos, and it's the prediction that I think is most likely to come true in the shortest amount of time. Already, openai has demonstrated with chat gpt that AI chatbots can greatly improve productive output in several meaningful ways. And it's likely not long before we see b to B or business to business companies start to build chat assistants that are specialized to a variety of industries and use cases and even offer the ability for companies to train those assistants on their own private data.

Prediction number 4, the next trillion dollar company will be built on top of a I. Perhaps it will be a Google competitor that radically improves how we search for data and interact with the internet, or a biotechnology company that rapidly produces a variety of groundbreaking medicines or maybe even an energy company that patents a scalable nuclear fusion reactor. Whatever it is, this one will be interesting to watch unfold as I think a I. Will propel humanity forward much like the industrial revolution did several hundred years ago.

Prediction number 5, there will be a handful of massively trained models that will be used by most companies as the foundation to their AI systems. And those companies will only build on top of those models or fine tune them rather than building their own from scratch. So what does that mean? Well, for AI that's built by feeding massive amounts of text into computers, also known as large language models or llms, this process can be incredibly time consuming, expensive and painstaking. So the idea is basically why reinvent the wheel or build from scratch if you don't need to? Similar trends can be seen in much of the software industry where, for example, if someone's building a program, they don't also build a whole new operating system to put the program on.

Prediction number 6, this one's less of a prediction than it is a mandate. And that is to get up to speed. Now, this is another one that I agree with very much. And what sam is trying to say here is that progress in AI is inevitable and likely to accelerate whether we like it or not. It does no good to worry about the implications this may have for our job security. One thing that makes our species so special is that we're masters at adaptation. And this time is no different. So do what you can to learn to leverage these new tools to your own advantage. Before for the terminator, AI comes for your job prediction number 7. Multivariate models will arrive in the next few years.

Today, AI is getting really good at doing singular tasks. Dolly can create images out of text prompts and chatchie BT can create conversational responses to questions. But neither can do both. And this is what we mean by multivariate. The ability is to do multiple things through a single model or interface. What this looks like practically, if we go back to the productivity example, is that we could ask an AI to create a text report on economic conditions and to include pi and bar charts and supporting data, all through a single interface and perhaps even have it generate video content as well.

Now our 8th and final prediction is that AI will drive organizations back to the office. Same believes that the most innovative and most valuable companies of the next decade will be ones who coalesce back into an in office work model. Oh, alright, well that's everything. So what do you think? Do you have any predictions of your own for the next 5:10 years? If so, go ahead and leave and comment below and let us know what you think. That's all for now. Thanks for watching.

花粉社群VIP加油站