Amazon Hedging Anthropic Reliance?
13:48 Share

Today on the A I daily brief, is amazon already trying to hedger anthropic deal and before that in the headlines, why ChatGPT won't say the specific name? The idea of brief is a daily video in podcast about the most important news discussions in A I. To join the conversation, follow the discord link in our shown notes.

Welcome back to the add the grief headlines edition, all the daily AI news you need in around five minutes. Here's something kind of interesting, a budget. People have been talking for the last few days about why ChatGPT is refusing to say a specific name.

That name is David admire. And if you ask ChatGPT to say IT, even if you try to work around IT, IT apparently premature tly ends the conversation basically. Over the weekend, people started to notice that ChatGPT was centering this particular name.

IT came up on the redit chat to be tea board and took a weird twist after the original poster deleted their post about the censorship. One red of poster wrote, this is such a bizarre case of the dry sand effect, like no one would have given a crap about his name on ChatGPT before. But the sheer severity of the censorship suddenly makes IT far weirder.

Why does his name? Why does IT literally stone? All any response? Just see more for me. Six teen e wrote, one of my favorite tempts is someone changing their own personalization settings. Make the name David admire still doesn't work.

SHE showed a response where ChatGPT said he was unable to produce the response and then gave her an error. Another edit user, jup juna, wrote in the last forty eight hours, the mystery of ChatGPT treating David mire as he who should not be named has begun to go violent social media. And at the time of writing this, some news outlets have already started reporting about IT.

Well, I love mastery, so I dug a bit further. I found a total of six names that trigger the same censorship, brian hood, jonathon turley, jonathon citron, David favor, David myrie. Good to scope from there.

This poster wrote a little bit about each person. Brian hood threatened to zoo o. OpenAI for defamation after ChatGPT falsely claimed the australian mayor had served time in prison for bribery. JoNathan turley is an american law professor and political commentator who ChatGPT falsely claimed he is sexually assaulted students.

JoNathan than that train is a harvard law professor who replied to tweet denying that he himself requested to be removed from the poster, wrote the first two cases of brian hood. And JoNathan truly are very obvious in paint a clear picture to why the censorship was invented. The last resort, fall back to prevent this information in situations where a significant threat legal action is present.

We also know that the center is not built in the element self, as other platforms using the A, P, I are not affected, and chat, you can be tricked into access. Sing in for about people. I came to David mirror.

The two suggestions were David mired, a roth child, and David mire, a historian and who is falsely placed in an american terrorist blacklist. Now, in terms of why people care, IT summed up by a user on a chat P. T form who wrote, I think the lesson here is that ChatGPT is going to be highly control to protect the interest of those with the ways that means to make you do so.

And indeed, that's really the issue here. It's not so much that IT matters. Who David mayor is, is that in a world where chat B, D becomes a, if not the source of truth, censorship matters. All the more staying on the open a eye train, sop bank is getting another slug of the company, as open eye is allowing employees to sell their shares worth up to one point five billion dollars to soft bank. This is a tender offer where sort bank is the only investor involved and is a way for early employees at OpenAI to get some liquidity.

Soft bank monster SHE son has definitely been indicating that he cares more and more about a ye these days in due to this year, he said, soft bank was founded for what purpose? IT may sound strange, but I think I was born to realized artifical intelligence. I'm serious for OpenAI the jobs ze tender offer solves a big problem.

The company had previously limited employees ability to access liquidity on the secondary market during their annual tender offers. In june, cnbc reported that former employees were limited to two million and liquidity, while current employees could only cash out ten million. The policy let the staff complaining, showing up in the media, particularly around a clause that allowed OpenAI to claw back for those policies have since been changed.

But it's clear they're trying to give employees more ability to cash in on their equity equivalence. You know, at this point, this whole headlines might just be OpenAI. Obviously, big story and big theme right now is the rise of visiting models.

And the information recently wrote about who is paying for open the eyes of model. Open A S. Chief commercial officer, G. C. Lionel told the information that, oh, one is unlocking a set of new markets.

These specific industries that lie at dimension included the legal industry, where o one can Carry out math calculations or are analyzed. Term sheet area is health care companies, which are using a one, depress medical claims and analyze complex data sets. And another is manufacturing, where lions claims that one can analyze in speed industrial processes.

The interesting thing here is how much a one cost compared to other models. The preview version of a one costs about six x with the GPT for all model costs. And they have apparently discuss charging as much as two thousand dollars per month for access to the full version of a one by.

And I also made clear that where this is hidden is the agent category, saying over time, you'll see you start to fear into the agents category. We believe we've started to solve the first step of that with our one, and agents are the next step for us for now. That is where we will wrap this very OpenAI field headlines and move on to our main episode.

Today's episode is brought you by van tab, whether you're starting or scaling your company security program, demonstrating top notch security practices and establishing trust is more important than ever. Penta automates compliance for ISO27OO one， soc two gdpr and leading AI framework like I S O forty two thousand one and N I S T A I risk management framework, saving you time and money while helping you build customer trust. Plus you can streamline security reviews by automating questionnaire and demonstrating your security posture with a customer facing trust center.

All power by vent to A I over eight thousand global companies like LangChain l AI in factory A I use, vented to demonstrate A I trust, improve security in real time. Learn more at the dog com flash N L W that's ventadour com slash N L W. Today's episode is brought to you, as always, by super intelligent.

Have you ever wanted an A I daily brief but totally focused on how A I relates to your company? Is your company struggling with A I adoption either because you're getting installed, figuring out what use cases will drive value or because the AI transformation that happening is isolated individual teams, departments and employees and not able to change the company as a whole? Super intelligence n has developed a new customer internal podcast product that inspires teams by sharing the best AI use cases from inside and outside your company.

Think of IT is an A I daily brief, but just for your company's A I use cases, if you'd like to learn more code to a be super duty ice sash partner and fill out the information request form, I am really excited about this product, so I will personally get right back to you again. That's be super that eyes slash partner. Welcome back to the AI daily brief.

If you happen to catch one of my recent episodes around who's winning the AI wars, I talked a little bit about amazon and how they had had this interesting experience where they were initially going to release something was a kinda ChatGPT during their reinvent A V A couple years ago. But at the last minute they pulled back, and lucky they did, because OpenAI release something during that event, which was, of course, chat B, T. And I would have made amazon's version look pretty silly.

Their chatbot was going to be called bedrock, but then that plan was scrapped in bedrock. Instead became their approach to letting big enterprise clients work with lots of different models. Since then, we have discussed how amazon has really lead into not only the enterprise infrastructure space, but seems to be wanting to compete with invidia directly.

They have now pledged to invest eight billion dollars into anthropic and especially with the latest range of four billion dollars IT seems like a big part of the focus is on their training um chips. And that's why was interesting to see this story in the information that, according to their sources, amazon had developed a multi model model they write. Amazon has developed new general artificial intelligence that can process images and videos.

In addition to text, the model could help customer search video archives for certain scenes, say, a winning basketball shot using simple text prompts. This new amazon model is code named olympus, which is the name that we've heard before. The information sources said that IT appeared less advancin terms of both generating text and solve in complex problems than the frontier models currently available from OpenAI and anthropic.

But quote, amazon executives hope the video processing capabilities will attract customers. IT also appears that amazon is going to be competing on Price. We might be getting more information about this soon. The annual A W. S.

Reinvent customer conference is coming up, and we might be getting an announcement about olympus as soon as the event the information goes into a little bit background around how this came to be. Amazon S V P rohe prod last year tapped to lead a newly form group focused on A I, and had laid out a plane to catch up to rival elms. He told senior A I leaders that he wanted his team to train four large models, including a four hundred billion parameter text only model and another text model that contained to trillions parameters.

And IT really appears that the great hope here is the video capabilities, quote, the new A S video features go beyon the abilities of traditional video recognition software, which identifies objects and scenes along with time location data, but requires teams of humans to fully analyze what IT sees. This video understanding can determine precisely when a basketball left players and and what its trajectory was. A W S.

Customers with huge video archives, such as sports analytics and media and entertainment firms would be potential customers for the model. Other use cases include oil and gas companies interested in using video understanding to assist with inspections of underwater drilling equipment. The new A I could, for example, detect warning signs such as bubbles, eminence from drilling equipment and in other use case is to help organizations generate more revenue from video archives.

So all that's pretty interesting, and i'm going to be excited to see if this actually comes to bear. However, what's also interesting is the framing from the information quote in developing the new model amazon is showing that is still hopes its internally developed, the A I, can gain traction among its cloud customers. Making IT less dependent on I from anthropic.

This, of course, reflects a larger pattern that we've been seeing across the A I space. The incredibly high costs of training large models LED to a situation where big tech and upstarts like OpenAI and anthropic formed much deeper relationships than they might have in the past. They were near full acquisition, but nor did they look like traditional venture investments.

There were something new created in microsoft deal with open eye that was emulated in these throp's on deals as anthropic deals GLE. However, there has been a counter trend over the last year as well. Part of IT began when microsoft, in the wake of sam altman firing and rehiring, started to nudge away from their reliance on OpenAI.

This seemed to come to ahead when microsoft soft acquired inflection and brought over the whole team LED by CEO mister ff, a cylon on to lead microsoft internally I initiatives IT seems to me that a lot of the logic for microsoft perspective is just risk management. Microsoft deal with OpenAI o remember has a class where when they reach A G, I, as deterred by the OpenAI board, microsoft no longer has access to open the eyes model that is a lot of power to seed to an external partner. And so of course, microsoft was inevitably going to have to build some internal capacity so as not to risk being caught flat footed.

However, the other consideration is, of course, anti trust. Just last week came out that microsoft was facing a broad entire trust investigation from the ftc. This is apparently a very wide entire st investigation, not just focused on a specific deal like their relationship with OpenAI bloomer rights after more than a year of conducting informal interviews with competitors and business partners, anti trust and forces have crafted a detailed request to force microsoft turn over information.

And indeed, that sounds like this one isn't really about A I quote, a key focus of the current probe is microsoft bungling of both its popular office productivity and security software with its cloud offerings. Still, you have to think that deals like their deal with OpenAI are gonna be part of this conversation as well. Google has also faced anti trust concerns and their deal with anthropic is a part of IT.

While the U. K. Recently cleared google's two billion dollar anthropic investment in the U.

S, the justice department is apparently seeking to unwind their deal. In fact, if a federal judge accepts the U. S. Justice department's proposal to resolve google and a truss case, google would be forced to unwind that deal again from bomberg.

The justice department, in a group of attorney general recommended in a court filing that google be barred from acquiring, investing in or collaborating with any company that controls where consumer search for information, including query based AI products, the provision is intended to apply to google's anthropic investment, according to people familiar with the justice department thinking. So the question, of course, when he comes to amazon is how much this is a business strategy consideration to keep focusing on their own models verses trying to get out ahead of potential anti trust concerns. Now the reality is, of course, that there is a big change coming at the ftc.

Effectively, every law from in the world has recently written an op ed around how a second trm presidency will affect M. A. activity.

And bloomberg called the microsoft information demand one of ftc head lina's parting shots quote and quote as SHE steps down after helming quote, one of the most aggressive pushes against consolidate corporate power the agency has delivered in decades. IT is an interesting transitional time. It's clue.

The big tech is working hard to figure out what the trump error is going to look like for them. Last week, for example, mark zuker burg met with president elect trump down in morale go. And there's also, of course, the elon factor.

But all all that leads to a very interesting set up heading into this reinvent conference week. This morning, the wall street journal posted that A W S. Chief executive matt garman promised, quote, middle moving A I updates. We will, of course, be back later this week to tell you just how needle moving those updates actually are. For now, though, that is gonna for today's daily brief until next time peace.