The Information's TITV
00:25:23 · Joelle Pineau, Akash Pasricha
[00:25:23] S3: First and foremost, we're really committed to our modeling roadmap, the ability to build frontier models. And in some cases, it's very specific capabilities. In the past, we've focused on multilingual capabilities. Now we have a lot of efforts around cybersecurity space as well as enterprise use cases, which can be quite different from some of the consumer AI. And so a lot of that infrastructure, obviously, to continue to have a really ambitious roadmap in terms of training larger models, generic synthetic data to train those models. But there's also definitely an appetite as AI grows to adopt our solutions for small medium businesses who may want to lean on us for compute and also, you know, to provide a greater number of inference solutions. So those are some of the things that will certainly come into the picture as we go. Right now, we've shared actually is our commitment to build some of that compute in Canada, which is part of the government of Canada's involvement in this funding round. [00:26:30] S0: So tell me about your outlook then on the many different options for chips that you have available to you that are coming online? I'm sure you're using NVIDIA. Are you using chips other than NVIDIA?
00:26:30 · Akash Pasricha, Joelle Pineau
[00:26:30] S0: So tell me about your outlook then on the many different options for chips that you have available to you that are coming online? I'm sure you're using NVIDIA. Are you using chips other than NVIDIA? [00:26:46] S3: We are definitely using NVIDIA chip. I would say that's the bulk of it. We've also had some really good results working with AMD chips. And beyond that, I would say a lot of what we're doing is looking at the possibilities. We've, you know, we've considered using TPUs and others in the past. And so we're always, you know, for us, it's really about benchmarking the performance of that. Between training and inference, the needs can be a little bit different also. And then we increasingly are generating more and more synthetic environments, synthetic data. And so that in some cases opens up the possibilities. [00:27:19] S0: What about some of these smaller inference focused chips, you know, SambaNova, Cerebras, d-Matrix, you're talking to any of them? [00:27:27] S3: We are definitely talking to them. I wouldn't say we have huge amount of our workflows on them right now, but we are talking to them or keeping an eye on what's going on over there, definitely. [00:27:38] S0: Right. And, I mean, you talked about the growth of the company as well in the announcement. So where is the run rate at right now for the company?
00:27:38 · Akash Pasricha, Joelle Pineau, Ken Gawrelski, Nikhil Raghuveera
[00:27:38] S0: Right. And, I mean, you talked about the growth of the company as well in the announcement. So where is the run rate at right now for the company? [00:27:47] S3: I mean, the the company in terms of, you know, how we're we're growing is is very much focused on our global strategy and in terms of integrating our new acquisitions, alafalfa, great partnership with with German company. Also, few smaller companies recently announced Aisle as a new partner for Cohere based in Prague, expertise in cybersecurity, Reliant AI, which had an expertise pharma. And so a lot of the growth in the last few months is about broadening either in terms of like the global markets we are going to go after in the case of our partnership with alafalfa, but also specific industry verticals, domains where we think there's a lot of appetite for the work that we're doing. [00:28:38] S2: Right. [00:28:38] S3: In the next case, you know, financial services, pharma, cybersecurity being really top of mind for us. [00:28:44] S0: I'm just trying to get a sense for for the the multiple here on 5,000,000,000. What is the revenue right now for the company? [00:28:52] S3: I mean, we should get our our CFO Francois here online with you to give you all the numbers. I I [00:28:58] S4: I would love
00:28:58 · Nikhil Raghuveera, Joelle Pineau, Akash Pasricha
[00:28:58] S4: I would love [00:28:58] S3: to. I mean, is... Is the latest ones. All I say is, you know, like, quarter over quarter, and we just closed our q three. We've been meeting our we've been meeting our objectives. We've have a really healthy revenue stream and and and talking to lots of other prospective customers. And so I'm not worried per se in terms of our ability to meet our revenue objectives in the coming year. [00:29:21] S0: Right, right. Hey, so kind of a fun moment for all AI labs around the world to see the Anthropic S1 come out. I'm sure you poured over it like we all did. Tell me about what surprised you there and what implications you felt like it had for your business as you build it. [00:29:44] S3: Yeah. There's a lot of information in there, and certainly it it's [00:29:47] S0: 400 pages. Compared
00:29:47 · Akash Pasricha, Joelle Pineau
[00:29:47] S0: 400 pages. Compared [00:29:49] S3: to some of the the other ones we've we've seen in the past. You know, some of the things that are really catching our our attention is is navigating how to talk about the safety and the risks of these systems. You know, I've been building systems from research point of view for many, many years. We're used to obviously like benchmarking these systems and looking at what are some of the risks that could come out of that. The disclosure of these risks in this case is a very interesting one with a number of things we probably hadn't seen before. And so it's going to be curious to see whether that becomes standard across AI companies' IPO, whether that we adopt a certain form of how to report on things, in many cases that are quite speculative, quite frankly, how the market reacts to that. So I think that's the that's the thing that I'm I'm looking at, in particular deciding how do we build up the right body of knowledge so we can be transparent about risks in a way that inform the investors while being grounded really like the science and the properties of the system. [00:30:54] S0: When is Cohere's IPO going to be? [00:30:57] S3: I will be glad to give you a call as soon as I have the date. I would love to come back and tell you all about it. [00:31:03] S0: You just got just got to tell Francois to join us next time so that we can... [00:31:09] S3: We'll come together. Always a pleasure.
00:31:09 · Joelle Pineau, Akash Pasricha
[00:31:09] S3: We'll come together. Always a pleasure. [00:31:13] S0: One last question before you go, Joelle. Just going back to the compute infrastructure and capacity conversation. So as you look to build out your own capacity, I mean, we've seen all this conversation around Neo Clouds and the risks maybe that those businesses face because of the fact that as companies like yourself, the hyperscalers, as they build their own facilities and as they come online, will they move their capacity from the Neo Clouds, and what will that mean for their businesses? Walk me through just your own perspective on that. Do you see that as reality for you guys too? Are you using NeoClouds right now, and are you likely to transfer that? Because maybe that can tell us something about the future of these NeoClouds altogether.