The AI Daily Brief: Artificial Intelligence News and Analysis - The 15 Most Important AI Products of 2024

Episode Date: December 30, 2024

It was a year of incredible AI innovations and beloved products. In this episode, NLW counts down the most significant. Will it be a coding tool? Research platform Perplexity? Share your favorites in the comments. Brought to you by: Vanta - Simplify compliance - ⁠⁠⁠⁠⁠⁠⁠https://vanta.com/nlw 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 we are counting down the 15 most important AI products from 2024. 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. Hello, friends. We are here in our end of year episodes, and today we are counting down the 15 most important AI products of the year. A couple caveats before we get into this. One, this is just my opinion, although it is not the most important products to me. it is my ranking for the most important products in general.

I'm taking into account, obviously, all the time that I spend preparing this show, as well as all the conversations I have with big companies as part of Super Intelligent, plus my own experience to come up with this list. Another caveat, I guarantee the second I am done recording, I'm going to think of something that I forgot. And third, by the very subjective nature of this type of thing, I'm sure that I'll disagree with myself before long. In terms of criteria, most important is obviously a little bit vague,

but I'm thinking about things not only like how used, they were, but how influential, how they shape the conversation, and how they pretend for where we're going in the future. We're going to start off with a couple of honorable or frankly dishonorable mentions. Firstly, 2024 was 100% absolutely not the year of AI wearables. This is probably best summed up in Marquez Brownlee's review of the Humane Pin, which he titled the worst product I've ever reviewed. For now, there were also other products like the Rabbit R1 that didn't really see traction, and just overall, the combination of capabilities, form factors, and everything else was not there for AI wearables this year. I don't think that this means we should count AI

wearables out far from it. I'm glad that these companies are continuing to try and build interesting novel experiences. I'm optimistic for the future. The future just didn't come in 2024. My second dishonorable mention is Apple Intelligence. After years of waiting to understand what Apple's Gen AI strategy was going to be, we finally started to get information that was shared at the Worldwide Developer Conference in June, and then from there we got delays and completely underwhelming features. The idea behind it, trying to find AI applications that actually blend into the background and just make people's lives better, feels both pertinent and very Apple. But so far, they have completely missed a mark on that, and Apple heads into 2025

with some serious challenges. Not only have they not gotten people to upgrade, for the first time many of us are actively thinking about leaving. From there, though, we're going to move into our actual top 15, where we get into things that were actually meaningful and important. At number 15, I have a tie between 11 Labs and HeyGen. 11 Labs is an AI audio platform, while HeyGen is a video avatar company. The reason that these are on the list but so low is reflective of the fact that A, these are incredible products, which are incredibly useful right now, but B, they haven't really fully made

their way into mainstream business usage yet. I'm fairly confident that's going to change, and 11 Labs in particular has been pushing hard to be a valuable partner for AI agent creation, so I'm expecting a lot from these companies in the new year. At number 14, we have Devin. When Devin was first demoed back in the first part of the year, people were absolutely blown away. It basically felt like the promise of having an actual junior developer at your hands that could build things from the ground up. What made Devin appealing is that it wasn't just a coding assistant, but something that could go even farther. Now, in practice, Devin has only just become available. It's expensive at $500 a month, although, as some have pointed out, that works out to about $8

per hour of coding time. And in its production version, it is still being positioned more as a complement to existing developers rather than as a replacement or supplementation. The reason I have it on this list is really just as a flag that one of the big unlocks from AI is going to be when true no-code tools bring the power of software creation to non-developers. I think we're getting a glimpse of that with Devin, even if it's not fully there yet. At number 13, we have the first example of a particular type of cheating.

I'm going to do a number of times on this list, which is bundling a set of companies. And basically, the companies that I'm focused on here are specifically designed enterprise products that actually are going into production and showing enterprises that agents aren't a far-away thing, but can actually be helpful now. Sierra, Decagon, and Ada, all went into actual usage as customer service agents in 2024. And I think it's incredibly important for enterprises to have that first-taste type of experience

in a low-hanging area where agents can really perform to get more confident about agents in the future. Given how important I think agents are going to be as a theme in the coming years, I think it's worth a mention. Right now there isn't one breakout leader yet. All of these companies are being seriously considered and used by big enterprises. And while it's slightly different, I also wanted to give Glean a shout out here. A lot of companies have tried to be the be all end all enterprise or work AI type platform. And Glein is really the only one that I've seen start to break through it all. Whether that can hold and this incredibly diverse set of solutions and products really make sense all in one bundle

versus a separate relationships for enterprises remains to be seen. but there's no doubt that in my conversations with enterprises, these guys are getting some traction. At number 12, we have another combo, Venice AI and Elon Musk's Grock, and both of these I'm putting on the list for being loud proponents of at least much more uncensored AI. Venice is particularly mission-oriented around uncensored AI, creating an interface that plugs in a number of different models, and also focuses on privacy, while GROC and XAI is sort of the one mainstream frontier lab that clearly has a strong belief set around this as well.

I think as LLMs become more and more the way that we mediate knowledge in the world, having private by default and more uncensored approaches is going to be really important. At number 11, we have another combo pack. It's Fathom, Reed, Fireflies, Otter, and frankly all the other meeting note takers. When you dig in with enterprises, if there is one ubiquitous use case, it is meeting summarizers. Now, I've talked a lot recently around how I still think we're just actually. figuring out how to get value out of these things, but there is no doubt that they are basically the single most default, can't imagine before we had it type of use case inside the enterprise.

One shout out here, maybe a little forward looking. Granola is a slightly different take on something similar, trying to have a more intelligent and really minimalist approach to this meeting summarization, and the people who are using it are loving it. And I think Granola reflects a trend which we're going to talk about a couple of times today, where part of the story of 2024 was user experience becoming more and more important. So Grinola maybe wanted to keep an eye on going into 2025. At number 10, we have Suno and Udio. Two companies that together showed that music generation was viable and got artists really starting to get excited, if also potentially a little bit nervous. In point of fact, though, by and large, I think musicians are viewing these things as tools in their

toolkit. And part of why these made the list this year is that we actually started to see them infiltrate the mainstream. The best example of this was BBL Drizzi. This was originally created by artist and comedian King Wallonius using AI, but got super famous when it was sampled by Metro Boomin on a disc track against Drake that went uber viral. It's been a fairly quiet second half of the year for music generation software, but I would anticipate hearing a lot more from these companies in 2025. Now we are in the top 10 friends and we kick it off at number nine with Lama 3.1 405B. Since the leak of the original Lama's model weights back in the beginning of 2023, meta has been the standard bearer and biggest frontier lab champion of open source AI. The reason that I'm calling

out Lama 3.1 405B specifically is that this was the first open source model that was truly GPT4 class and really close the gap with closed source. The fact that we now have open source models that are very nearly as performant as the absolute state of the art is incredibly important in determining how the industry is likely to evolve. And so I think that this model is very deserving of a spot on this list. At number eight, we have ChatGBT's advanced voice mode with vision. This is interesting because from an announcement standpoint, it's about seven months old, but from an actual we've been able to use at standpoint, it's very, very new, just actually becoming available at the beginning of December. I think, in fact, that this is a paradigm shift

in how multimodal LLMs interact with our world that will have a huge impact on how we use AI in the future. I think that it's probably only number eight on this list because we haven't really had a chance to get used to it yet, but my guess is that it fundamentally changes how we interact with chat GPT. And of course, now that Google's models have similar capabilities, this is quickly becoming table stakes. I for one know that I haven't fully adapted to this new vision paradigm, and I'm really excited to see how it changes my usage in the year to come. At number seven, we have a nod to the future. I'm giving this slot to Salesforce's agent force. And there are a couple reasons why. It's not that agent force is somehow radically breathtaking in terms of what it can do,

but what it is is an actual agenic platform that's trying to have software due labor, not just assist with it, that's in production and live and available for sale from a major company in Salesforce experimenting with a new business model. Remember, the way that this platform works is that you pay based on interactions, not based on seats. And so it really is at the dead center of a huge number of trends that I think are going to be enormously important in the years to come. And Salesforce is out there trying to claim ground now. I would expect that our 2025 list a year from now is going to have a lot of agent presence. And so in some ways, it feels like agent force is a preview of the future. At number six, we have another category prize, which is the video generation category V-O-2,

Sora, Luma Labs Dream Machine, Runway, and PICA. Interestingly, you could make arguments for each of these individual products as being the one to be highlighted. The argument for V-O-2 is that it really does seem to be the farthest to long and have the best sense of physics, but then again, it's only brand new, and we haven't had much of a chance to play with it yet. Sora is also only just recently fully available, but the announcement of Sora in February created a huge, huge amount of excitement and momentum in the video generation space. When it comes to these other companies, I think they all should be recognized for taking advantage of the momentum that had been created by the SORA announcement, and actually making tools that were available for people to use while

that open AI model wasn't available. PICA could also be highlighted for its focus on social media, and Runway could be highlighted for its partnership with big Hollywood studios. It's also worth calling out Kling, which represented a set of Chinese AI video models that for a time seemed to be really outperforming their American competitors. I think still NetNet video generation had a much smaller role in the year than many of us might have thought that it would, but we close the year with it kind of catching up with itself, and a whole new set of use cases, I think, are unlocked heading into next year. Today's episode is brought to you by Vanta.

Whether you're starting or scaling your company's security program, demonstrating top-notch security practices, and establishing trust is more important than ever. Venta automates compliance for ISO-2701, SOC2, GDPR, and leading AI frameworks like ISO-402, and NIST AI Risk Management Framework, saving you time and money while, helping you build customer trust. Plus, you can streamline security reviews by automating questionnaires and demonstrating your security posture with a customer-facing trust center, all powered by Vanta AI. Over 8,000 global companies like Langchain, Lila AI, and factory AI use Vanta to demonstrate

AI trust and prove security in real time. Learn more at vanta.com slash NLW. That's vanta.com slash NLW. If there is one thing that's clear about AI in 2025, it's that the agents are coming. vertical agents by industry horizontal agent platforms, agents per function. If you are running a large enterprise, you will be experimenting with agents next year. And given how new this is, all of us are going to be back in pilot mode. That's why Super Intelligent is offering a new product for the beginning of this year.

It's an agent readiness and opportunity audit. Over the course of a couple quick weeks, we dig in with your team to understand what type of agents make sense for you to test, what type of infrastructure support you need to be ready, and to ultimately come away with a set of actionable recommendations that get you prepared to figure out how agents can transform your business. If you are interested in the agent readiness and opportunity audit, reach out directly to me, NLW at B-Super.A.I. Put the word agent in the subject line so I know what you're talking about,

and let's have you be a leader in the most dynamic part of the AI market. All right, now we are moving into the top five in the realm of the heavy hitters. At number five, and frankly, this one might be too low in retrospect. respect, we have OpenAI's O1 model. The big theme for the last part of this year has been the plateau that we seem to be hitting when it comes to pre-training and the scaling models that Frontier Labs have been using for the last couple of years. O1, I think, represents a branch on the evolutionary tree of LLMs, where a new set of strategies focused on improved reasoning, in this case test time compute and giving the model time to think at the point of inference,

are starting to be tested as ways to get more performance out of our LLMs. The different right now between 01 as a reasoning model and GPT-40 and its ilk is a little fuzzy. It's not even as though 01 is clearly better at everything. There are many things that it is not as good at. But it appears, at least from where we're sitting, that 01 is the start of where the next generation of models are headed. And in that way, might be hugely significant to how the industry evolves. At number four, we have perplexity. And there are a few reasons why I have this so high. First of all, Perplexity is just one of the most beloved products

from an actual consumer standpoint. Those who use it use it all the time. It's tended to replace Google search for them in many cases. It's opened up new types of research. And I think it showed that creating a great product was just as important as having a good model powering it underneath. This was, I think, one of the major themes of 2024,

the resurgence of product in U.X thinking, and perplexity was a major reason why that happened. Plus, you have to give the company credit for being honestly the first credible challenger to Google Search pretty much since Google Search emerged. And while the company has an incredibly hard road ahead of it, it's one that I think many of us are rooting for from the standpoint of being loyal pretty much everyday users. At number three, we have Claude. Anthropic had undeniably a really good year. 3.5 Sonnet was for much of the year, the default model for many. It remains the default coding model for many users. From an enterprise business standpoint, Anthropic doubled their share of the

market from 12 to 24%, according to Menlo's recent Enterprise AI report. But the reason that they're so high on my list is that same theme that I put around perplexity, which is that this is a company that drove product thinking across the entire AI space. They brought in a co-founder of Instagram as a product leader, and it almost immediately started to show. Around the middle of the year, they launched artifacts, which is basically just a better interface for using LLLN. Once you used artifacts, it was extremely hard to go back to the very simple chat-type interface of chat GPT and other LLMs. And perhaps as the best indication of how good it was, it was basically copied by OpenAI

and Google and everyone else as a more standard user experience for LLMs. Anthropic has continued to push Clod to be a better and better product experience, and given how many people's behavior either switched to Clod this year or integrated Claude in some way, I think it deserves its spot here in the top three. And lastly, we come to the end of the list, and this one is tough, mostly because I think there's a really strong argument for both of these products. It may be, in fact, that it's only my bias as a non-coder that leads to the sequencing being what it is. But at number two, I have cursor. No product was more ubiquitous in its particular segment than was cursor. There were lots of

great AI coding tools. Repplet had a great year that had tons of people building with them. Devin, I mentioned earlier in this list got people really excited, but Cursor was absolutely everywhere. Developers voted with their feet, and a huge number of them moved their entire workflow to Cursor, and that is why they are number two on this list. And given how important coding is as the initial breakout use case of generative AI, I think Cursor has to get the nod this high up on the list. And as I mentioned, it might only be my bias as a non-coder that kept it from the number one slot. For those of you who are regular listeners of the show, you will know exactly what the number one slot is then. The number one most important AI product of the year was Google's Notebook LM.

And of course, where Notebook LM specifically came to life was the audio overviews. Notebook LM had actually been around for a while before everyone started paying attention. It was the introduction of audio overviews, which would take any set of sources or documents and turn them into an audio podcast conversation between two people that really started to gather people's attention. Over the last few months, audio overviews have just gotten better. First, Google introduced the ability to better steer the outputs, and more recently, they've even further expanded the ability for people to interact with those hosts as a part of the generation. Notebook LM also now has an enterprise version, and that's a good thing, because in our conversations

it's super intelligent with enterprises, no product has more captured people's imagination than notebook LM. And this is why it gets my number one slot. Notebook LM does not fit in the category of one-to-one replacement for things we already do. Instead, it opens up totally different and new opportunities. I think that over the course of the next couple of school years, it will become absolutely de rigour and default to start new topic exploration with the Notebook LM podcast. Already, whenever I want to, quote-unquote, read a new dense research paper, you better believe it goes straight into audio overviews first. Businesses are exploring how to use Notebook LM-style audio overviews to summarize and capture knowledge and present it in a way that's much more

accessible. Dealing with email inbox overload, sum it up with Notebook LM. Want to get a sense of what customers are saying? Feed all the customer service transcripts into Notebook LM and get an audio overview. Every day on Twitter, I see some new use case like this one where Andre Carpathy said that he wants to read books alongside Notebook LM. Carpathy was one of the first to really get this. He tweeted back in September, it's possible that Notebook LM podcast episode generation is touching on a whole new territory of highly compelling LLM product formats. Feels reminiscent of ChatGBTBT. Maybe I'm overreacting?

A couple months on from that, I don't think he was overreacting. And that is why Notebook LM is my AI product of the year. Long episode, maybe the longest non-interview in AI Daily Brief History. But a very fun one. Hope you enjoyed it as much as I did. Appreciate you listening as always. And until next time, peace.