We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
People
A
Axel Springer CEO兼出版人
A
Axel Springer全球战略合作伙伴关系高级副总裁
O
Omar (AI企业家)
O
OpenAI
O
OpenAI应用研究主管
P
Perplexity CEO Ari Zehavi
R
Roman Chung (The Rundown)
S
Sam Altman
领导 OpenAI 实现 AGI 和超智能,重新定义 AI 发展路径,并推动 AI 技术的商业化和应用。
亚马逊CEO安迪·贾西
商学院教授
微软CEO萨蒂亚·纳德拉
投资者
播客主持人
播客主持人,专注于英语学习和金融话题讨论,组织了英语学习营,并深入探讨了比特币和美元的关系。
教授
Topics
亚马逊CEO安迪·贾西:亚马逊计划增强Alexa的功能,使其不仅能回答问题,还能执行任务,这体现了亚马逊在AI领域的战略布局和技术实力。然而,项目的进展缓慢,且存在潜在风险,这需要亚马逊谨慎应对。 播客主持人:对大型科技公司财报的分析显示,华尔街对AI的反应褒贬不一。谷歌的云收入强劲增长,且内部代码有四分之一由AI生成,获得了投资者的积极回应。Meta的大规模投资尚未带来足够的回报,这引起了投资者的担忧。微软的云计算部门增长强劲,并从AI中获益良多。苹果公司虽然营收创历史新高,但AI相关的产品并未刺激销售增长。 OpenAI:ChatGPT搜索旨在快速直接地回答问题并提供来源引用,这与Perplexity等其他AI搜索引擎形成竞争。OpenAI与新闻出版商等有许可协议,以确保结果来源可靠。ChatGPT搜索目前仅对部分用户开放,未来将向更多用户推出。 Axel Springer CEO兼出版人:AI搜索将成为未来获取信息的主要方式。与OpenAI合作将为新闻业带来新的商业模式。 用户评价:ChatGPT搜索速度快,来源标记比Perplexity更好,但深度不如Perplexity Pro;ChatGPT搜索有时会产生幻觉,准确性有待提高。 Sam Altman:计算能力的不足限制了OpenAI的发展,OpenAI正在努力解决计算能力不足的问题,但一些产品的发布因此延迟。 Perplexity CEO Ari Zehavi:Perplexity优先处理信息查询,并改进网页导航功能,这体现了Perplexity在AI搜索领域的创新和竞争策略。

Deep Dive

Chapters
Amazon is rearchitecting Alexa to make it more action-oriented, leveraging AI to take actions beyond just answering questions and summarizing data.
  • Amazon is planning to reveal new foundation models for Alexa in the near future.
  • The new build, internally code-named 'Remarkable Alexa,' will reportedly cost between $5 and $10 per month.
  • Amazon faces higher stakes with Alexa due to its widespread use in families and kids.

Shownotes Transcript

Translations:
中文

Today, in the AI daily brief search, GPT is now live, and the A I searchers are heating up before that, in the headlines, amazon is getting agents and other news from hyper scale earning season. The AI daily brief is a daily podcast in video about the most important news and discussions in A I to join the conversation fall that described in our show notes.

We are fully in earnings season right now and have a bunch of stories from that as part of our headlines today kicking off, amazon is looking to give agency to alexa. During yesterday's earnings call, amazon C E, O and Jesse hinted a gentle version of the company's virtual. He said, I think the next generation of these assistance in generate A I applications will be Better at not just answering questions in summarizing inie and aggregating data, but also taking actions.

And you can imagine us being pretty good at that with alex a Jesse confirmed that efforts are still on going to quote rearchitects, the brain of alexa, a process that includes a new set foundation models that the company is planning to reveal in the near future average to build an a eye power elect, first revealed back in twenty twenty three with prototypes using model supplied by anthropic. In practice, we haven't seen much of this situation of the project, say, for an unreleased version that struggle to turn on smart lights and was painfully slow. The new build, which is internally code named to remarkable alex a, which is so painfully crying, will reportedly cost between five and ten dollars per month alongside a strip back free version.

IT was originally slated for an october released by bloomberg reports that the time line has shifted into next year. Earlier this week, bloomer chronical some of the difficulties that have held back the launch of an eye e power alexa. They noted that amazon is facing a fundamentally different problem than other A I labs.

Bloomer rights users toying with chat B T, expected to make mistakes. If amazon turned on its aleem brains and the legs, I began spewing provocative advices, IT could turn into a fiasco. Jc, given the huge portion of kids and families that use echo hardware.

And this is a really interesting counterpoint to the fact that these big tech companies have such big distribute. Yes, alexa has a huge number of built in devices, but the stakes are also higher. The expectation and how accurate A I has to be is just higher when IT comes with device darties installed LED and integrated into your life.

And from some new thing that you're just trying to figure out how much to trust that or not. Overall, big tech earnings this week revealed a lot about how wall street is feeling about a eye. All of the public hyped scales have now reported quarterly ly financials.

And the response for somebody of a mixed bag, google saw a positive response earlier in the week with investors liking strong growth in cloud revenue and and darci statement that one quarter of internal code is now being generated by artificial intelligence. On wednesday, meta and microsoft each saw negative responses for their own reasons. For meta, the story continues to be about massive investment, with two little to show.

For IT, the company is set to hit a hundred billion dollars in costs across all divisions this year, around six billion going to AI research and infrastructure. The company is generally showing solid growth, with quarterly revenue at a record high forty billion dollars, up nineteen percent from this time last year, slightly slower than the twenty two percent growth they posted last quarter, but still a solid pace. The escalating investment cost is what starting to concern investors.

Remember, meta is just coming out of a period of intense belt tightening. The company was criticized for overspending on the virtual reality division, which was such a large bet that IT, of course, prompted a name change that division has struggled to deliver a return on investment and is still posting over four billion dollars in quarterly losses. ZARA berg was force to face the music and lay twenty twenty two, signaling a year of efficiency, his words conducting multiple rounds of layoffs, cutting way back on investment spending.

That error is clearly over, with soaker c stating during the earnings call, R, A, I, investments continue to require serious infrastructure, and I expect to continue investing significantly there. The company guided, quote, a significant celerity and inflections expense growth next year as we recognize higher growth appreciation and Operating expenses of our expanded infrastructure fleet. Meta is somewhat unique among the hyper scales and that they don't have a cloud division that is clearly ripping the bending of A I adoption.

Instead, the company needs to show a return on this massive spent through metrix like time on platform and advertising and efficiency. They did attribute eight percent increase in engagement time on facebook and six percent on instagram to A I powered feds, but investors are still skeptical. Ryan o barrie's.

ax. Investment management said there's been an enormous amount of capital spent on developing ai, looking very specifically to see who is monetizing IT, who's actually giving a return on investment and what does that look like. Microsoft, meanwhile, is in a slightly different position.

The company showed strong growth in their cloud division, which is capturing the lion share of the eyes, gains the group group by thirty three percent far of expectations. Copecks for the quarter were almost fifteen billion, which came in above expectations and was a fifty percent time from last year. CEO soften the della said.

Our AI businesses on practice are passed in annual revenue runs rate of ten million dollars next corner, which will make IT the fastest business in our history to reach the smile stone. Still, the company also admitted that azure cloud is basically a capacity, and the company needs to construct more data centers to expand. Apple was a little lackluster with the stock falling in after hours trading.

They reported revenue at all time high, prepelicza modest rebound, and iphone sales grow by six percent. But the problem is that apple doesn't see a big dump in sales coming. They guided load amid single digit sales growth for the december quarter, and these were supposed to be the two quarters where apple demonstrates that A I can drive an upgrade cycle.

So far, consumers have not responded to apple intelligence is must have feature. But as we talk about a number of times, features are barely there so far. One thing to note is that with all of these companies, the result to are extremely positive by traditional metrics.

Basically, the bigger shift is what wall street expects and how much is about real rally now versus the promises of future returns. Still, as much as wall street is in small ways punishing companies for high capital expenditure, it's pretty clear whether over investment or under investment is worse. A sisal, the president and CEO benioff capital management said the CEO of a tech company is much more likely to get fired for not spending enough on A I S.

Spending too much because right now that's the deciding factor. And IT, was that going to do for today's a daily brief headlined edition? Next up, the main episode, today's episode is brought to you by Venus. Venus is a private, uncensored, generate A I APP IT access is open source models to enable text image in and co generation without the fear of being spied on or having your data exploded.

Discuss anything with Venus without concern about IT being monitored, sold or given to advertisers in governments, Venus is different because your conversations and creations are kept security within the browser, never stored or accessible by Venus, unlike other AI apps. Then I won't tell you what's OK to say or not, then is want to patronized you IT simply provides direct access to machine intelligence. No topics or off limits, no ideas or taboo.

With Venus, your in control of the A. I, as you should be. Prosued scripts are available for forty nine dollars a year, or eight dollars per month.

Daily brief listeners receive a twenty percent discount on Venus pro, visit vena ice lash N L W and enter the discount code N L W daily brief, that's N L W daily brief, all one word. Today's episode is brought you by super intelligent. Every single business workflow in function is being remade and reimagine with artificial intelligence.

There is a huge chAllenge, however, of going from the potential of a to actually capturing the value. And that gap is what super intelligence is dedicated to filling suu r. Intelligent accelerates AI adoption and engagement to help teams actually use A I to increase productivity and drive business value.

An interactive A I use case registry gives your company full visibility into how people are using artificial intelligence right now. Pare that with capabilities, building content in the form of tutorials, learning paths and a use case library. And super intelligence helps people inside your company is show how they're getting value out of ai while providing resources for people to put that inspiration into action.

The next three teams that sign up with one hundred or more seats are going to get free. Embedded consulting. That's a process by which our super intelligence team sits with your organization, the specific use cases that matter most to you, and helps actually ensure support for adoption of those use cases to drive real value.

Go to be super diet. I to learn more about this. A I enable ment network. And now back to the show today, we are discussing opening eyes ChatGPT search.

But really what we're discussing is the absolute search wars that have now engulfed the internet. IT has been a very long time since anyone actually, actually, practically threaten google dominance of search. And yet now IT is so clear that whether it's google that wins or not, what search means is going to change any fundamental way.

We've got perplexity out here innovating, really pushing the boundaries of what the summary model of search can be. We've got google trying to catch up with their overviews and at the same time trying to figure out how they can do that without disrupting their own business model. And now we have OpenAI entering the ring as well.

In an announcement post, open eye said that their new chat P T search service is designed to provide fast, timely answers you would have previously needed to go to a search engine. And for, in other words, the idea here is that rather than having a cruise around and click on links and try to find the right reference point, you instead just get a summary that answers the question right there for you. OpenAI s model is powered by a fine in version of the GPT four model, and can show up photos and information from the internet, open a eye, gave examples of sports scores, news and still quotes, and just like with perplexity, users can also ask follow up questions to refine their search.

ChatGPT search produces both inline inside bar attribution to news publishers in other dataset. Ces OpenAI has licensing deals with, and this is an important point. One of the things that OpenAI keeps saying as they sign these deals is that is not about training data is about how they surface the most reputable sources in their search.

For example, results related quarries about the election will be directed to sources like the ap and reuters service is now available to ChatGPT plus and team members on mobile and the web, with other subscribers getting access over the coming weeks, followed by a rollout to free users. The company has also released a browser extension to allow users to make chat P, T, searched their default search engine. OpenAI said they planned to keep improving search, particularly in the domains of shopping and travel.

They also planned to add voice mode and access the reasoning models for what they call deeper research. Speaking to the idea that this is where search is headed, a quote from the e drive, as the C E O and publisher of amount, we are convinced that A I search will be in the near future for the next generations. A primary way to access information.

Partnering with OpenAI positions amount the forefront of the shift IT allows us to test innovations at an early stage while safeguarding journalism, core values and integrity. But the essentials, the S V P of global strategy partnerships that EXO spring says as A I reshaped the media landscape, axel springer partnership with up in A I opens up tremendous opportunities for innovative advancements. Together, we are driving new business models that ensure journalism remains both trustworthy and profitable.

I do not think it's at all an accident. OpenAI is going to pains to highlight all of their partnerships with companies like the associated press, contain the dash married financial times and set up, especially given that complexity is now starting to be a wash in lawsuit and trying to catch up on their own licensing deals. Opening up for their part, are very clearly excited about this same and tweed, hey, i'm really started to keep piping our own product, but you really should get chat, P, T, plus, install the chrome extension for search.

I am carefully the first to admit when we ship something that isn't very good, but this time it's really good. On the positive side, some people noticed that IT was really fast. A, I entrepreneurial, omar said, my only question about search GPT, how on earth is IT so fast? Seriously, what and how did they cook? It's nearly instant.

Open the eyes, head of applied research. For as power responded, feedback was that IT slows, we put effort into making IT super fast, to which sully said, teach me away, roman chung, from the run down rights. First impression, trying IT feels fast.

They d even pull our logo. Second impressions y continues around two exceptio than perplexity pro, but not as in depth. Rone continues. I like the source tagging Better than perplexity, says the company name rather than a number.

No related questions at the bottom, like plexus, which is one of my favorite plexi features roman then gives us power rankings, my AI sert rankings as of today, perplexity pro over P, T. search. ChatGPT search over a google search.

Google search over bing. This doesn't mean ChatGPT search. Google or meta can eventually beat replete ity, but IT does mean the AI web search.

French just got serious, but what of caution? However, he finally adds, ChatGPT search is hu Cindy. When you ask IT for tweet, x is notorious for being hard for search engines.

Perplexity at least says, I can see content instead of making something up. And that need brings us to some of the negative comments. Hu cino are very high, among them professor of the molecule ghz, finding very high hlubis ation rates right now.

I'm sure IT will improve investors, Stephen oski says, made up stuff, non reproducible, answers slow. What more could you not want from search, copy and business school professor cristian hendricks and rights? I have searched GPT and my very first search that was a failure.

Later in the conversation, IT started completely polluted ating. The newest dev diary plexus is still might go to Benjamin in to cracker, who IT should be noted, is that organize competitive X A I. Now I don't really get ChatGPT search, tried multiple and quote searches, and the results are very similar, tragic, Normal GPT results, very text heavy, was expecting a more search first experience.

Now it's clear, as you heard from all of these things, is that the perceived battle is with perplexity. Perplexity is in an absolute arms race to advanced features. CEO are from trinobantes tweet yesterday until now.

We mainly prioritized informational queries. Search is about anything you want to do. Navigational queried is essentially a link in a site map. As an answer, we made IT even easier to navigate the web, especially useful when you set perplexity as you're defauts search, develop nec doboys response. This is huge.

The only reason I didn't have replete xy said, as my default with the twenty five percent of time I needed to click the top result and navigate to testing catalogue news rights, perplexity just got even faster. But what super important is, is accurate perplexity one search GPT zero. Overall, the sentiment is pretty positive.

Allum rights open a eyes design team is really stepping up and crushing ing with their OS experience. GPT search looks incredibly slick to and with their strong ground position, their well place to capture a massive slice of the market probably the most. And indeed, we're kind of dealing with layers of distribution here.

Perplexity recently posted of serving ten million monthly active users, which is amazing for how fast are growing. The same time, that number is closer to a couple hundred million for ChatGPT, giving them undeniably an edge on the distribution front. And of course, there's google who serving up something like eight billion inquiries a day, which means you have to think that if they get IT together, they have such an inherent advantage going in.

Regardless, that is going to be a super interesting battle, one that is hard to imagine. We don't benefit from its consumers. One more little piece of OpenAI news before we get out of here today.

An interesting sub story from a redit ama with OpenAI C O M. yesterday. All man says that the lack of computer is holding the company back, he wrote.

All of these models have gotten quite complex. We also face a lot of limitations and hard decisions about how we allocate our computers. Many great ideas.

Of course, we ve seen a wave of headlines over recent months about how open a eye is trying to address this issue. On the training side, the company is partners with oracle to build one of the largest training clusters in the world in texas. For inference, they recently announce that would be using M D G P U and have plans to start designing and manufacturing their own chips.

Partners with road common T S M C. Now, as an example of a feature being held up by a lack of compute, often refer to vision capabilities for ChatGPT s advances voice mode, stating that IT wouldn't arrive any time soon. This feature was showcase during the demo in April but didn't arrive on the rest of voice mode was rolled out in late september.

Other big news from the A, F, A included other indefinite timelines on new products, an update to OpenAIce, generate image model delhi has no timely without stating we don't have a release plan yet sora, the company's generated video model has also been held back, with Kevin wheel, the companies 的 chief product officer, saying the delays due to the quote need to perfect the model, get safety in personation other things right, and scale compute. Although the ama was filled with news about delays, altman insisted we have some very good releases coming later this year. Nothing that we are going to call GPT five. There you have that that is the story from open eye, and that will be at the end of our daily brief today. Appreciate your listings is always and until next time kiss.