The AI Daily Brief: Artificial Intelligence News and Analysis - Financial Data Shows Business AI Usage is Way Up

Episode Date: August 14, 2024

NLW covers a recent study from Ramp that uses proprietary card spend data to show how usage of AI is growing in the enterprise. Plus the latest on ChatGPT's latest model. Concerned about being spied on? Tired of censored responses? AI Daily Brief listeners receive a 20% discount on Venice Pro. Visit ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://venice.ai/nlw ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠and enter the discount code NLWDAILYBRIEF. Learn how to use AI with the world's biggest library of fun and useful tutorials: https://besuper.ai/ Use code 'podcast' for 50% off your first month. The AI Daily Brief helps you understand the most important news and discussions in AI. Subscribe to the podcast version of The AI Daily Brief wherever you listen: https://pod.link/1680633614 Subscribe to the newsletter: https://aidailybrief.beehiiv.com/ Join our Discord: https://bit.ly/aibreakdown

Today on the AI Daily Brief, financial data shows that AI usage in business is way, way up. Before that, in the headlines, we have a new model in chat GPT. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes. Welcome back to the AI Daily Brief Headlines edition, all the daily AI news you need in around five minutes. As you have heard over the last week or so from my show, there's been a lot of speculation about a potential new model release. It turns out that at least as of the time of recording, which to be sure is before when some leakers had suggested that a bigger new model would be announced, OpenAI has confirmed that something new has been going on. Yesterday evening, they tweeted there's a new GPT4-O model out in chat GPT since last week.

Hope you're all enjoying it and check it out if you haven't. We think you'll like it. Now, I had mentioned that some people had suggested that based on their anecdotal evidence, their personal experience, there did seem to be something improved. Hyper Right CEO Matt Schumer, for example, wrote, Something might be going on with GPT-40. For the first time, in a long time, it provided better vibes on an output than 3.5 sonnet. Really surprised. We'll keep using it today to see if it continues. Others pointed out that GPT40 seemed to be able to actually beat certain tests that it had struggled with in the past, such as answering which number is greater, 9.11 or 9.9.9.

Yet, interestingly, not everyone was all that enthused with this announcement, And it wasn't just because of some potential disappointment that it wasn't a bigger release, although there was certainly a bit of that as well. Some people shared examples of answers where it still struggled, at least one of which the official ChatchipT Twitter account responded to saying lots of work left to do or on it. But there was some consternation about the nature of the release in general. Professor Ethan Mollick writes, another release without release notes. I get the point also made by Google about their mysterious new model that it is hard to write

release notes for LLMs. But I would love to see some attempt to explain what changed. People actually use ChatGTPT for real tasks and need guidance. Gradyo founder Abu Bakr Abed gave an example of a context where this is a real problem. They quote tweeted the ChatGPT announcement and said, This is why academic research should never use LLMs via APIs. You have no idea what model and surrounding software you are interacting with and it can be changed at any time without notice.

Might as well throw reproducibility out the window. Accelerate Harder writes, actually annoyed by this. Due to randomness and confirmation bias, people always try to claim ChatGPT changed when it hasn't. but now they are actually updating it without telling anyone so these speculations will never end. Now, for my part, I am always open to and excited about a new model. I haven't had a chance to dig in too much to see if I've noticed improved performance, although I was impressed with Chad Chapit's ability to design a custom Magic the Gathering set based on HP Lovecraft over the weekend.

So maybe that was the new model in action. But I still do also think it's fascinating that we really continue to be just making these tiny incremental updates. As I've said before on the show, this is either because those are the only updates that are available right now, or because there's a larger sea change lurking, and because of the competitive landscape, OpenAI doesn't yet feel the need to jump all the way to those increased capacities. There could also be safety issues, et cetera, but in either case, for now, these sort of incremental upgrades seem to be all that we're going to get. For completeness, we should mention speaking

of incremental upgrades, that over the last couple days, Google has been making moves as well. They announced a Gemini 1.5 Flash upgrade that not only promised improved output, but did so at 70% reduced prices. They also rolled out fine-tuning to all developers and added support for 100 plus new languages in the API. Now, when it comes to new models that are exciting people, there have been a bunch of announcements. One announcement that's getting some attention is this one about an AI model that has the ability to detect diabetes, stroke, COVID, and other diseases with 98% accuracy by studying a human's tongue. The new system in question was developed by researchers at Middle Technical University and the University of South Australia in Australia, and claims to have

the ability to diagnose a wide array of conditions. Said senior study author Ali Al-Nazi, typically people with diabetes have a yellow tongue, cancer patients and purple tongue with a thick, greasy coating, and acute stroke patients present with an unusually shaped red tongue. Key features of this evaluation include the color of the tongue, shade of the coating, form of the tongue, depth of the coating, oral moisture, tongue crevices, contusions, red spots, and tooth impressions. Now interestingly, this particular model was designed by replicating a 2,000-year-old technique from traditional Chinese medicine. Speaking of AI and scientific research, there is a ton of buzz on X right now about Sakana

AI. As summarized by GROC, Sakana AI, in collaboration with the University of Oxford, has unveiled the AI scientist, an advanced AI system engineered to fully automate scientific research. The system is capable of generating research ideas, writing code, conducting experiments, summarizing results, and even writing and peer reviewing entire scientific papers. Operating at a cost of about $15 per paper, the AI scientist aims to accelerate scientific potentially reducing the need for large human research teams and exploring a wider range of hypotheses. This innovation marks a significant step towards integrating AI and advanced scientific knowledge, possibly leading to breakthroughs in AGI.

Now, the responses to this are really a Rorschach test for how people think about AI safety right now. Many people are incredibly excited about what they see as the potential for a new era of scientific discovery. Others are a little bit more nervous that this could lead to a so-called fast takeoff, where AI starts improving self-recursively in a way that we can no longer control. Lastly, today, AI continues to find its way into the real world with the Tokyo Metropolitan Government, launching an AI system that uses high-altitude cameras to detect fires and building collapses in real time to accelerate disaster response during earthquakes. Given that experts believe there is a 70% chance of a significant earthquake occurring

directly beneath Tokyo within the next 30 years, this particular use of the technology is highly non-theoretical. That, however, is going to do it for today's AI Daily Brief Headlines edition. Next up, the main episode. Today's episode is brought to you by Venice. The leading AI companies store your entire conversation history and attach it to your identity forever. That's every question you ask, every answer you receive, every image you generate, every thought you share with the machine it's all being spied on. If you trust all the companies, hackers, and NSA board members that will ever have access to your AI conversations, then rejoice, for you are well served.

For the rest of us, Venice is an alternative. Venice is a powerful AI app for text, image, and code generation that respects you as a sovereign individual and believes privacy, and free speech are not only human rights, but necessary for civilizational advancement. Private, permissionless, and uncensored, you can try it for free without an account. AIA Daily Brief listeners receive a 20% discount on Venice Pro. Visit venice.a.l. slash NLW and enter the discount code, NLW Daily Brief. That's NLW Daily Brief. All one word. Today's episode is brought to you by Plum. Generative AI promises to supercharge your

productivity and give you superpowers, but if you're not an engineer, trying to harness AI can be incredibly frustrating. Hours wasted wrestling with complex tools only to give up when they don't work. We all have tedious tasks we'd love to automate and challenges AI could solve, but few of us have the skills to fully leverage these game-changing technologies.

That's where Plum comes in. The mission? To make automating your work feel like magic. Imagine typing out AI, read my Gmail and ping me in Slack when something critical comes in, and watching it come to life before your eyes. No coding required. Whether you're a marketer, salesperson, or founder, Plum enables you to create custom AI workflows

in minutes, not hours. Check out useplum.com, that's Plum with a B, for early access to the future of workflow automation. Today's episode is brought to you by Super Intelligent, the platform that helps teams maximize AI. Super is, of course, the platform that we've been building that pairs fun, fast, practical video tutorials with step-by-step instructions to get you actually using AI, and from there unlocks information about hundreds of use cases that show you how people are actually getting value out of AI right now. Now, we have just launched Super for Teams. This is a new add-on experience that allows teams to share more information about what AI they're using, how it's working, and how to get more value out of it.

Whether your company is 25 or 2,500, superintelligent is going to be the best platform for unlocking information about how to get the most out of AI right now. If you'd like to learn more about the super intelligent teams offering, go to Bsuper.A.I. slash partner and send us a note so that our team can get right back to you. Once again, that's B-Supert.a-I slash partner. Welcome back to the AI Daily Brief. One of the big discussions on this show is, of course, the discussion of how much AI is actually being adopted. You will frequently hear me, quote, skeptical op-eds, and then refute them with the evidence that we have based on our experience with super-intelligent. But anecdotes versus anecdotes is always a hard argument to sort through. And so I'm

always thrilled when we get some actual factual data. The data that we have today comes from Ramp. Ramp calls themselves the ultimate platform for modern finance teams, and basically it's a company that helps other companies with their team spending. They have corporate credit cards, spending limits, approval flows, a vendor payment system, and more. In other words, for the purposes of this conversation, they have a lot of data about how companies are spending and on what. And as you'll see, information on how companies are spending on AI. Yesterday, they published their 2024 business spending benchmark summer report, and one of the areas they focused was on artificial intelligence. The company writes, ramp processes billions of dollars each month,

giving us a front row seat for spotting new tools that companies are testing as well as vendors that are becoming mainstream. We've grown our database of AI vendors to over 1,500 vendors, including long-time tools that may not have started as AI focused but have since incorporated cutting-edge AI capabilities. Our latest analysis of Ramp customers aggregated and anonymized transactions shows AI was the fastest growing expensive Q2, as more companies lean on the technology to unlock efficiency gains and new product capabilities. So let's discuss some of their key findings. One, they write, companies are making bigger and longer term AI investments. Longer duration in prepaid services they write tend to go on AP. And in Q2, mean AP spend for AI vendors rose

375% year over year. Also, it seems that retention is up. With ramp writing, customers are staying with top AI vendors. 70.4% of customers who started spending with these vendors in 2023 continued to spend after 12 months. Second key finding, Foundation AI models top the list of first-time software purchases. OpenAI leads the list of software vendors that companies purchased for the first time in the first half of 2024 followed closely by Anthropic, suggesting companies are actively tailoring models for internal operations and customer-facing capabilities. Third, companies are purchasing software to scale creative functions. Tools for generating images, audio, text, and video they write are gaining traction. Grammarly, Mid-Journey, and 11 labs all

rank high on the list of top AI vendors. And finally, while this isn't exactly related to AI, it does tell a story that includes AI. Key finding four, businesses are supplementing their workforce with independent contractors. Despite a job market favoring employers, businesses of all size boosted spending on flexible labor solutions like Upwork. Combining these purchases with AI spending, companies are taking advantage of options to boost productivity without hiring new staff. So let's come back and look at a few of the key nuggets of information or insight derived from this. As part of the release of this report, they put together a blog post called Companies Go All In on AI. Five charts show how. Chart one relates to AI vendor stickiness. Ramp writes last year, AI was

was still an experimental area for many organizations. Very few companies spent on a recurring basis with vendors we categorized as AI. The amount they spent with these vendors on ramp cards and ramp bill pay was also nominal. Now they say businesses are going all in. Our data show customers who started spending with top AI vendors in 2023 are likelyer to stick with these vendors than those who started spending in 2022 or 2021, a signal that companies are increasingly convinced of AI's value. And here's the comparison of businesses that started spending with top AI vendors in 2022. Just 42% were still spending with them a year later, whereas those who started with AI vendors last year, 70% are still spending with the same vendors 12 months later.

What's more, these companies are spending more. The mean accounts payable spend with AI vendors has nearly tripled since the start of the year. year. The trend of longer duration AI engagements is also evident if we look at AP spending trends for AI vendors. In Q2, mean AP spend with AI vendors rose 375% year over year. The number was at around 20K in quarter 2, 2023, getting all the way up to almost 100K in quarter 2 of 2024. There was a huge jump between quarter 4, 20203 and quarter 1, 2024, going from around 35k to around 90K. They also said that these AP expenditures were dominated by OpenAI, with company spending $100,000, 881,000 on average. Even as certain solutions become more locked in, however, it also appears that

there's a lot more experimentation. The third chart that they share is that card spend with AI vendors doubled year over year, which they call a sign of further experimentation. Quote, if AP spend represents companies making significant vendor commitments, then card spend indicates companies' willingness to explore different vendors or increase usage of less expensive tools. Typically, companies pay for new tools month to month via credit wherever possible and only shift to AP for annual subscriptions. Mean card spend with AI vendors increased 104% year over year, suggesting companies continue to experiment even as they start to lock in their favorite tools. In quarter two of last year, the mean card spend with AI vendors was around $1,000, which has gone up to a little over $2,000 in quarter

two of this year. In terms of specific companies, Anthropic is apparently doing really well. Right now, they say machine learning operations and platforms are the clear winners of the AI category, with nearly 80% of customers that are using AI transacting in that category. Anthropic, they say, is seeing particularly rapid adoption when we look at the percentage of businesses purchasing AI models on ramp cards. Anthropics market share was around 4% at the start of this year, but jumped to 17% in quarter two. In July alone, they say the number of transacting businesses stored 22% month over month, and total card spend nearly doubled. This is some of the first data evidence that we've seen that the introduction of Claude 3.5 Sonnet, as well as new product features like artifacts,

are clearly manifesting themselves in Anthropics performance. The last chart that they share is the fastest growing AI vendors by customer count. They sum up by saying AI productivity tools and open AI alternatives are growing fast. Part of what this chart shows, they say, is more evidence of Anthropics growth, but they also write that vendors that offer novel approaches to everyday work are becoming commonplace, as companies turn to AI to boost employee productivity. Suno AI, for instance, lets anyone generate music using simple text prompts. Limitless runs in the background to capture your audio and screen automatically so you can generate meeting, summaries, emails, and more at a later time. Instantly offers AI for sales engagement and lead

intelligence. Curser makes an AI code editor. And you can see all of these tools showing up in the top 10 of fastest growing AI vendors by customer count. And that's across small, large, and mid-market businesses. Ramp co-founder Eric Gleeman really sums this up when he tweets, AI is no longer a toy. This really is some of the best evidence that we have so far of just how much adoption is increasing. This is not survey data. This is not self-reporting. This is actual factual spend. And it is increasing with speed. Really interesting stuff. Great job to ramp on publishing this. I would love to see this from other companies like Mercury and Brex as well to help fill out the picture of just how AI is getting adopted by businesses. For now though, that is going to do it for today's AI Daily Brief.

Until next time, peace.