In this week's #AIEcosystem Report, Acceleration Economy practitioner analyst Toni Witt reviews the latest in #artificialintelligence #innovation, #funding, and solutions: Google continues advancing its #Gemini LLM; Axelera AI gains funds for its edge AI computing for AI chips; a young startup develops a therapy app called Sonia (YC W24). Tune in: https://lnkd.in/gK-bM2vd
AI Ecosystem: Google’s Gemini; Axelera AI Raises $68M; Sonia’s Therapy App
Transcript
If you're an industry accelerator, a cloud or Elon platform, a SaaS company building new AI tools, or a partner bringing all these pieces together, whatever role you play, please consider joining us an awesome Texas for the AI Ecosystem Summit, which is running the first quarter of 2025. You can learn more at aiecosystemsummit.com. Welcome back everyone to the AI Ecosystem report by Acceleration. Economy and this weekly series, we help you get smart about AI innovation, funding and solutions in just a couple of minutes. So let's get right into it. So a little while ago, we covered Google Gemini, the new massive LM that Google was developing and how Gemini would really be implemented across their product portfolio and really power a lot of the product releases and updates that the company is rolling out in the next few years. And we'll actually, we've seen some action on that front and we want to give you really just a quick update on how what Gemini is really powering right now and what is Google actually offering in terms of AI products, AI features and AI services, as well as just recent product releases in AI. First off, to get on the same page here, Google started off is actually really just a family of large language models, but now it's really evolved into just a portfolio, of course, the models themselves, but also AI apps, AI products and services. And so focusing on that first piece, the models, they first, they come in a few sizes, Gemini Ultra, which is really the largest Gemini model, Gemini Pro, which is a slightly more lightweight version, Gemini Flash, which is a again, a step down. And then Gemini Nano, which is actually two small, two smaller models, nano one and then nano two. And these are both actually meant to run on mobile devices. So purely on the edge and on that, this is actually a pretty cool mobile application. They launched on Gemini Nano, which lets you basically hit a button on your phone. And record speech and it auto transcribes and summarizes it, but right on your device. So there's no latency, there's no Wi-Fi required. You can be anywhere in the world, there's no data leaks because it's all happening on your phone. And all of these models are multimodal, unlike Google's other model Lambda, which is text only. Now Gemini apps are built on top of these models and at the base layer, of course, there's a chat interface similar to chat debt with these models and that's accessible through through web and mobile. And these Gemini apps, because they're multimodal, can accept PDFs, images, et cetera, right? Not just text. You can also import things files from Google Drive. And of course, it syncs across your devices that are logged into your G Suite account. Gemini is also working its way into, you know, Workspace apps like Google Docs, Slides Sheets and Google Search. Actually on search nowadays, I think this was a couple of months ago recently where like the first thing you see nowadays is if you type in a search query on Google, it comes up with an AI generated field at the top right. And Google Docs is a chat interface that you can engage with on the side that understands the document that you're working in in Slides, you can generate images automatically directly within Slides and so on there's. All features like this. Google Sheets also has some some small features around AI now. You can pay for extra I features like auto generating, for example, flight itineraries based on details like flight details that it pulls from your e-mail, like tickets, Google Maps history, if you're browsing around certain destinations or whatever. And these extra plans are all available through Google. One, there is a more expensive plan for really like the AI power users, if you will. And another cool thing is that now you can also create your own custom chat bots based on prompts. Like for example, you could say, hey, you're my private chef. Or meal planner, right? Give me a meal plan and then these like custom assistance. I can actually tap into various Google Apps to accomplish that goal, whether it's Google Calendar, Tasks, Gmail and more. And so just taking a step back, like I will say Google has had some misses in the past in terms of AI right barred was. You know, a questionable success and there has been some lashback on certain things with Gemini as well lately, but it really seems like the company is definitely not falling behind in the AI race like their features and these releases have been very powerful and value add to users. They're also in talks with Apple and other companies to bring Gemini into more environments. So definitely, you know, in the race definitely head on competition with Microsoft open AI, Meta and others. And yeah, you can always read more on their website about Gemini and. About these releases right now. Moving on to the funding section, a Dutch startup Accelera I just raised $68,000,000 in a series B funding round. This companies focus on the AI chip market in particular, which recently has gotten very crowded. Of course, those big players like NVIDIA, which is really the dominant player in this space. There's also been over $5.3 billion of VC backing in this industry as well, like for for chips and AI chips. So definitely competition, you know, is very fierce. Both on the startup side and on from these established players like NVIDIA. But we're accelerate stands out is really a, they're actually focused on a particular niche even within AI chips, which is actually edge AI computing. So when you want to run AI workloads closer to the edge, like on device, right, another is really a technical advantage that they have around using in memory compute, which basically increases the shifts performance by a lot. And obviously, I'm no like an engineer, so I can't really dive too far into this. But I will say the company is really, really deep into the research and technical innovation side of things. This company was actually born out of iMac, which is the Belgium based research lab that my dad actually works at. They have over 50 PHD's on the team. They're backers include Intestinal Deep Tech Fund, the European Innovation Council Fund, Samsung Catalyst. Samsung has a fund as well, and then a handful of others. The company is projecting nearly 70 billion in revenue by 2027. And you know, even a small slice of a huge pie like AI chips is relatively meaningful. You know, if they can dominate in the chips for AI, you know, edge compute market, they've definitely hit a home run. And overall, I'll say like they're definitely in for a challenge. You know, this is a very complex product. It's difficult to launch, difficult to sell. There's huge R&D costs, employee overheads. You know, you have these big players like NVIDIA and all these other AI chip startups from the US and Europe **** just ungodly amounts of capital into innovation in the sector. But at the same time, there's massive opportunity, right? If you get it right, You know, if you look at NVIDIA, they offer many, many, many things and there are huge company, the inherent advantage of of being a startup. Like this one is you can move fast and pick out small things that a big company like NVIDIA does. And if you can do a little bit better in a small part of what they do, you know, you can build a serious business. And in this case, they've picked, you know, edgy I computing. So, you know, we'll see. We'll see what happens right now. Moving on to the solution of the week. Today's solution might be a little bit controversial because we're actually talking about a young startup that's building an AI therapist and this is called Sonia. It's an app basically with a chat interface that uses. And interact with on a clinical level and the app can engage in therapy sessions about common issues like sleep problems, depression, anxiety. But can also detect if the user is really in a bad mental state, state of mind and perhaps considering violence or self harm. And in these cases the app will actually route you to national hotlines. You can also choose between voice and text to engage with Sonia. They have Co developed the app with several psychologists and clinicians and they are pretty early stage. They have about 10,000 people on the platform. On the app right now, they raised a couple million, including from YC. The app costs about $20.00 a month or $200.00 a year, which is definitely much, much more affordable than a real therapist from, you know, for most people. Not to mention it's, you know, it's always available. It's very accessible. And in many places in the world and in the US as well, you know, people actually don't have like physical access to human therapists. You know, if you're living off the grid a little bit or it's just not feasible after work for you to make your way to a therapist's office, this app is, is, is much more accessible in that way. And I think also a lot of younger people might feel more comfortable with their device actually, versus being with the human therapists. But this is definitely work gets interesting, right? Like there have been studies showing that the human touch of a therapy session is actually as important, if not more important than the actual advice or content of the discussion. And of course, the app, which is an app on your phone doesn't have that human touch. But on the flip side, the company does showcase research and user stats that show the efficacy of their platform that, you know, you have these people using it and they are getting better from it. And like I said, if it really can unlock, you know, just the accessibility of therapy and. You know, having these like, difficult conversations to a lot of people that otherwise would have zero access to real therapy. That is definitely a big advantage as well. So, you know, there's also concerns around bias, you know, bad content generation, you know, the quality of the answers of Sonia have to be very, very great given users are, of course, pouring their most vulnerable, deep personal thoughts and feelings into this. You know, like data has to be secure for that reason as well. So, you know, it's exciting. It's a young company. I think there's a lot of potential. Can think of some people in my life who would definitely want to use this. So we'll, we'll see. We'll we'll keep you updated on on, you know, their progress. But anyways, that's all for this week. Thank you as always for listening and we will see you very soon.To view or add a comment, sign in
Co-founder at Sonia (YC W24) | ex MIT, ETH, McKinsey
2wThanks for the feature Toni Witt Acceleration Economy You almost pitched Sonia better than we do ourselves :)