The bigger question for me is , at what point does investing in higher capability general models will stop showing the ROI?
For example - What percentage of workflows require this new highly capable model? How much of it can be replaced with the software tooling around it? What I mean is if the software tooling can optimize the query over a few iterations does it get the same output as from a single shot high capability model query?
I have over the last two months switched to OpenAI from Claude since it’s been vastly better for my cases. Wonder if I see any difference in the quality of the output with this change
The move the current companies are missing is to have a legislation declaring ai as a person. Just like how corporations are people, if ai is a person then they don’t have to show number of people fired, can argues content creation and engagement is through the roof and so many other possibilities. /s
I am sure this is great but it requires some depth in that field to appreciate this. I don’t understand if what they developed is a tech to do a live scan of any person or just a detailed map of one brain.
I respect cynicism and questioning stuff but this is misplaced. You have to trust the government since they are potentially the least partisan source here. Yes the data can be misconstrued by legislators but the truth of the data cannot be in question. It’s healthy to question it but the solution is to require proof of non-sabotage. It takes a lot of money and resources to pull this data together. It’s compiled by organizations across the world and being the trustworthy anchor is the most efficient way to achieve this. With that the government agency has every incentive to be non partisan and operate with integrity.
I don’t know if the circular financing is a problem. NVIDIA is the best my name in town, any company has to spend on NVIDIA assets for their compute. Now that makes NVIDIA rich and so they don’t know what to do with their money. They are just propping up companies they find interesting
This is alarming. Very soon there will be no point driving because insurance is going to jump in and mandate strict rules around how to sit, hold the steering wheel and how I should be looking and the fun of driving will be gone. This is all converging towards autonomous driving without a steering wheel.
I don’t understand the argument here. The article doesn’t describe a collapse or the breadcrumbs for it. The only argument I can put together is companies hosting the open source models in house or use some service like Amazon that could potentially host them and so replace the frontier models. Data center and specifically infra to host llms is still the main sticking point given the security concerns about data going to china. The article doesn’t make these arguments coherently
Another thing I noticed is the llms perform efficiently/effectively only under the optimum circumstances. May be this just a Claude issue, but when the session goes on for very long the effectiveness drops drastically and I start getting bad answers. This is especially true for design and debugging. Wonder how that ladders up to the token usage
This is not a new realization, it’s known. By definition llms are great summarizers not inventors. The value is in having the knowledge readily accessible and to cross check and validate against what’s already known - a huge time saver. It’s not going to come with anything new. It need not even, the value is in making things time efficient.
Of course the llm companies justifying huge investments want to frame it as a technology to invent new things and that is where the disconnect is. They are over indexing on the promise but companies are figuring out that the value is in automation and applying/showing value for existing workflows. Claude focused on this aspect and hence so successful
8k is cheap if laundry is fully offloaded but will a regular consumer spend 8k on a device that is not proven? I guess there is a subset of consumers that this automatically targets/caters to.
This is quite brilliant. This will seamlessly integrate with every collaboration systems as long as they are able to “enroll” Claude as a user. Claude gains knowledge across systems and makes the ai enhancements for each app/system unnecessary. For example if slack creates something similar it’s less useful because Claude’s model works across any system/app. More so, anthropic can either charge for it separately or not , companies will pay in extra used tokens anyway.
The issue is the cost is not going to be a hindrance for companies that have gone all in on the AI development. They may still find it cheaper than hiring engineers and if needed they will layoff a few more.
The companies that did not yet jump on this bandwagon and are still evaluating will have a decision to make.
No matter what the AI companies are going to change their pricing strategy and it’s going to become a lot lot more expensive to use. I am just hoping the price stays like this until I am done with my big chunk of work
ChatGPT doesn’t allow private relay and hasn’t allowed it since launch may be. So it’s not always possible to not use them, of course now there is no need to use ChatGPT and I have just stopped and moved on from it
Gemini made me turn google off completely. There is no provision to not allow your chat data from being shared send for training - even if you pay. Clearly they don’t care about end users and I am guessing they are losing money on paid cunaumer tier as well. But that is not acceptable to me. I just turned off all Google services and finally switched to ddg as the default search engine.
What does cursor have? An ide and coding orchestration? They are using Claude or codex for llms, so they get acquired for their user base and tooling? Feels like a lot of money for that given Claude has the majority mindshare.
This needs atleast a 30b model or Mr higher and so for most folks it means purchasing a new machine. Given the ram costs this may be become prohibitive and a monthly subscription may feel better roi
What is the target to write for is the key aspect here. Not sure there is enough on the phone for developers to create new experiences. I fear all of them are going to try to automate everything on the phone for the user. Not sure what value that provides. May be I am overly skeptical , let’s see
As a tangent, I don’t understand where and why meta fits into the AI race. They did not get any mind share (consumers) from the llms so far, granted they started the open source side to this but the Chinese companies produce far better models and have essentially become the default for on device set up.
They have ai glasses and integration into instagram and facebook as the other avenues. I don’t see ai glasses as compelling yet, and don’t know how much more ad revenue or user engagement they can squeeze out with llms baked into the IG of FB flows. They are spending a lot and not seeing any returns. Am I wrong in being pessimistic about meta with AI?
The chatbots(ChatGPT, Claude et al) showed Apple exactly what can be done, the user base is already well primed. So this is a product definition done for them to execute. If done well they will be able to provide a much stronger integration into the day to day use cases than the chatbots, and can siphon off user time from them. This time around the end to end is easier with Apple Intelligence and more importantly llms doing the work Apple is floundering at. So I am hopeful, but I still see the os/app level integration as not enough in terms of functionality to make it a hit. The primary use case for llms is still conversations and search. Apple should be focusing on that aspect primarily and also add the os/app level integration as a bonus - as something only they can do. If they just do the latter, it will not be as much of a success. Let’s see how they execute.
EDIT: To provide meaningful chat functionality they have to either eat up the cost or charge a subscription for it. This will be first time they charge for Siri - a product that doesn’t garner any positive reviews. This gets even more interesting to watch
So space data centers are absolutely possible then. I heard a lot of skepticism about the feasibility but it looks like Google and Anthropic looked at SpaceX and trusted them to deliver on the promise and even signed deals worth billions.
Anthropic realized security and safety are their main value prop compared to the competition. Either mythos or anything else since seem purpose built to streamline the messaging. It’s good, am not complaining, but i wonder how much this is intended to showcase what Claude can do over using it as is
I am not sure of that, may be in Europe, cannot see the U.S. banning. May be some states but highly skeptical there. Better chance of it flopping or doesn’t get to a good enough threshold to matter.