Claude Chronicles: Unveiling Anthropic's ChatGPT Rival's Major Updates

0:00The wait is over. Dive into Audible's most anticipated collection, The Best of 2025. featuring top audiobooks, podcasts, and originals across all genres. Our editors have carefully curated this year's must-listens from brilliant hidden gems to the buzziest new releases. Every title in this collection has earned its spot. This is your go-to for the absolute best in 2025 audio entertainment. Whether you love thrillers, romance, or nonfiction, your next favorite listen awaits. Discover why there's more to imagine when you listen at audible.com slash best of the year. There is a big new update to Anthropic, which is one of the biggest large language models after chat GPT.

0:47So today on the podcast, we're going to be diving into what exactly is happening here and what this big update is. So Anthropic, which is the AI venture formed by actually former OpenAI executives, has recently rolled out Claude Instant 1.2. So this is an upgraded version of text generating model and this new version leverages the power of the recently unveiled Claude 2 which essentially has considerable improvements in areas like math coding reasoning and safety and this is a very big deal for a lot of reasons but let's go into them first off according to data shared by Anthropic Claude Instant 1.2 actually achieves a 58.7 percent score on a coding benchmark so in comparison its predecessor Claude Instant 1.1 got a 52.8 percent So, you know, that's from 52 or maybe roughly like 53 to 59.

1:38That's pretty significant. That's a really good improvement, especially on such a small update like this. On a set of mathematical questions, the new version outperformed the older. It got a score of around 87%. And before it was at around 81%. So again, we're seeing big improvements from a small update on math and on coding. Now, this is important because currently even OpenAI is struggling with these areas, and they're relying on essentially companies like Wolfgram and other ones to kind of help them with plugins on areas that they are struggling on. But Anthropic actually mentioned in a blog update that Claude Instant now produces more extended, well-structured responses and is notably better at adhering to formatting guidelines, you might give it.

2:25right so if you ask it to come up with bullet points or if you ask it to use a certain amount of sentences or paragraphs or however you want it to be formatted with headings it's actually better at following those formatting guidelines now so additionally instant 1.2 has enhanced its capabilities in quote extraction multilingual tasks and also answering queries this is like really a big update in my opinion on just core functionalities for something that is going from you know 1.1 to 1.2 right like it's essentially like a tenth but um it is pretty significant I think the risk of Claude Instant 1.2 hallucinating or producing irrelevant text has also been reduced.

3:01It's also been, I guess they kind of like fortified it against jailbreaking attempts, right? Trying to get it to say stuff it shouldn't. And for the, you know, for those that don't know what jailbreaking is, essentially this just refers to using prompts to essentially get it to, you know, tell you how to do things it shouldn't. like, you know, rob a bank or whatever, a various other illegal or things that it shouldn't do. So what's really interesting is that Claude Instant 1.2 actually has a context window identical in size to that of Claude 2. So that is the exact same, which is about 100 ,000 tokens.

3:37And this context window indicates the textual data and model essentially examines prior to generating further content. So to put this into perspective, Claude Instant 1.2 and Claude 2 can evaluate around 75 ,000 words. Compare that to ChatGPT, right? Like we all know there is the limits when we try to put in a certain amount of text and it says, hey, this is too much text. You know, I've like pasted in a whole article and said like, hey, find me like five interesting important bullet points of things that are happening in this article. But then it's like, hey, this article is too long. We can't do this, right?

4:09This is not going to happen on something like Claude Incident 1.2 or Claude 2. So 75 ,000 words is approximately the size of a classical book like The Great Gatsby, for example. And typically models with really expansive context windows have a much better continuity in ongoing dialogues. Historically, I think we've noted that Anthropik's kind of overarching goal is the development of a next generation AI, you know, self-learning algorithms. So I think the company really envisions this technology powering virtual assistants capable of tasks like responding to emails, conducting research, or even, you know, like creating art and literature.

4:49They're kind of focused on making sure it can do this. So while we've seen glimpses of these capabilities and things like GPT-4, I think the true potential of this kind of technology, I think is really vast and we're just scratching the surface on it. So I think it is pretty essential to note that Cloud Instant doesn't really embody this whole revolutionary algorithm. I think it's positioned to rival beginner level products from major players like OpenAI and things like Cohear, you know, AI21 Labs. All of them are kind of navigating the realm of AI-driven text and image generation. And I think that's kind of where they fit.

5:24Since its inception in 2021 under, I believe their CEO is Dario M.O.D. He's the ex-VP of research at OpenAI. Anthropic has actually raised$1.45 billion in funding. Absolutely gargantuan. And they have a massive valuation in the, I think it's a single digit billions, but it is quite, quite large. So I think these new numbers are really impressive. The company's projections indicate that they're going to require a massive$5 billion over the next two years to actually build out their dream chatbot the way they want. So obviously this is very, very intensive in spending. They've raised a lot of money.

6:03they're investing heavily they're bringing big updates but you know you got to know that this is not this isn't some small startup that is just throwing out little bits of money here and there they are they've raised a lot and they are heavily investing so when they make these big updates to their models you know they definitely come at a cost currently i think anthropic has secured i think they have collaborations with thousands of clients and partners um i think specifically they have one with quora which gives access to claude and claude instant on its subscription based AI tool, which is called Poe.

6:32And in addition, I think DuckDuckGo's Duck Assist feature uses them. And that currently helps people with search queries. And they are using Claude and also ChatGPT. So kind of interesting. They got both platforms on there. Those are ones are going to have to kind of battle it out. But I think with the also the Notion ecosystem, they're embedded there. And then I think Claude also kind of helps with the technical framework for the Notion AI, which essentially is an AI writing aid that kind of integrates into the Notion workspace. A lot of people were surprised by this. They thought Notion was going to partner up with OpenAI, but indeed they went with Anthropics Cloud.

7:07So I think there's some really cool updates coming. This is going to make it much more user-friendly, much better of a tool for users. And to be honest, it does have a lot of features that OpenAI and ChatGPT don't have. That 75 ,000 word context window, right? You can upload an entire PDF to Cloud that's 75 ,000 words and ask it questions about it. And it can very quickly scan through, read, and give you answers. that's amazing it's not something you're seeing with open ai now the thing about that is it definitely comes at a cost very very high server costs when you're doing that kind of compute and so you know the thing that's a little tricky there is they're obviously spending a lot more and giving you a lot more bang for your buck but it's hard to know how sustainable and how long that will last if they'll actually be able to offer that if they're you know that's profitable most likely i'd say it is not and that is why we're seeing you know open ai not scale that back and for someone like Anthropics, Claude, their user base is much smaller and so they can kind of spend heavier to get their users because they weren't the number one where someone like OpenAI may not be able to financially foot a bill like that because of how massive their user base already is and they're already the number one player.

8:12So it's gonna be interesting to see what other improvements and features are coming out of Anthropics, how they continue to try to compete with OpenAI. Overall, I think they're doing a good job and I think the thing to be really watching out for is how sustainable they are long-term.