This might be the most perfect gif for the situation
I don’t feel like the companies are concerned about it costing more right now as much as they are betting that it will be cheaper in the long run. The cost of labor isn’t unlikely to decrease drastically while the technology is likely to become cheaper.
While I would love to believe Microsoft is being burned by spending on AI, I think they don’t mind spending more now so long as they can trade the cost labor for the costs of technology and maintain similar productivity.
Feels like they hope this will be to white collar jobs what Uber was for taxi drivers. Current profitability isn’t really the goal as much as being able to reproduce similar outputs.
Even if AI someday does become profitable short term, total profits will still go down long-term, because all profit comes from human labor (or exploiting nature). All any technology ever does in capitalism is to replace human labor, thereby putting more pressure on the empirically proven tendency of profits to fall. Profit gains from technology can only ever be short term and relative to competition who hasn’t yet adapted the technology. Once everyone has, prices drop, adjusting to lower socially necessary labor time.
The only way for the billionaire class to keep profits flowing a bit longer at this point is to do what we already see them doing now: get rid of the free market by enforcing monopolies with captive markets, bonded labor and merging big capital with an increasingly violent and warring state apparatus: capitalism inevitably leads to fascism-imperialism every time.
AI lends itself to this because of the centralized nature of data centers, the already monopoly based business model of tech companies and the political power and influence those monopolies hold. On the other hand, there is some potential, if not revolutionary at least disruptive potential, in small scale, specialized, open source models that can be trained with fewer resources.
The only alternative road to fascism, of course, leads to communism.
It will be cheaper for China with their massive solar panel production capacity, not for us tho
Hahahahahhhahahahahaaha
g a s p
HhahahhahahahbahahhHahahahahahah
There are ways to make it cheaper. Starting with maybe not encouraging token-maxing.
Generally, unless you’re either a FOSS project or generating images/video, you have to be doing something very wrong to spend more on AI than on salaries.
Not really. LLMs are still completely unable to manage even medium scale architectures. At a corporate scale they’re literally just spending on trying to have the most context they can in the LLM. There’s no getting around it.
Yeah, the smarter way to use LLM-based agents is carefully defined tasks. Mozilla describes their vulnerability assessment processes in this blog post.
Mozilla describes the process they’ve used: building a harness that instructs a model to find a specific category of vulnerability on a specific interface, and then write up its findings. It’s a narrow enough context that the model gets specific instructions, and a simple definition of success, and it sets up many such tasks that can be fed into the existing process for verifying and triaging bugs. Note that the output for this LLM pipeline basically feeds into the same interface for accepting bug reports from the public, or from their human contributors within the project.
There’s a couple of takeaways here, too:
- This pipeline is model agnostic. Mozilla set it up before Mythos was released, and its description of other models (Opus 4.7, Codex) confirms that Mythos is better but not a true game changer. The ability to swap out other models provides some assurance that the work done to develop the pipeline will be useful when cheaper or better models come along, or when a model becomes unavailable (like when a provider decides a particular model is too expensive to run, or a provider goes under).
- The increase in automated output (and presumably automation-assisted contributions from the public) has given the humans more work to do. Automation in this context actually increases the demand for human labor.
- Other projects will need to develop their own custom pipelines, specific to their project, to get good results from LLM based agents.
There are ways to use these tools, but none of it really seems like a truly revolutionary/disruptive change to how large projects are managed.
Generating images is not that expensive, it’s surprisingly inexpensive.
Yeah, it’s counterintuitive because it’s a lot more work for a human to draw a picture (much less a photorealistic picture) than to write a few words, but human language grammar actually has a lot of strict rules that makes that stream of letters work as “valid” output, much less “decent” output that kinda matches the prompt/description. Transpose a pair of letters or even substitute a single letter (or token) and you’ve got an output that just doesn’t work, in a way that generated images don’t have to worry about.
