Setting aside that we're living in a universe that's full of (practically) deterministic processes built over probabilistic components (and which behave sufficiently reliably without any human in the loop), I think the specific failure mode you're citing is that there aren't enough gates and constraints applied to the processes you've seen.
LLMs can contribute quite reliably given very narrow prompts and short horizons (keeping turns low and context brief). If you chain a bunch of these narrow contributions together and define guardrails (structured outputs, online evals, other-llm-as-judge/jury, etc...) you can produce a very repeatable workflow that reliably delivers to defined service levels.
The obvious issue being - you've got to define the workflow and implement all the guardrails, not hope that the LLM will infer them during a session or a one-shot prompt.
> Also I like credit cards for the rewards, cash back or especially travel points. But also things like extended warranty coverage and other perks.
Much of this is funded by inflated interchange fees paid by merchants (and thus inflating the cost of goods for everyone). Ideally you would just pay for the added value you see (fraud protection, charge disputes, supplemental insurance) rather than those costs being externalized to other buyers.
"Serving ads" is just the commercial application. All of the infrastructure used to harness (coordinate? co-opt?) the attention of the many and direct it toward some desired aims will surely end up being a tremendously transformative social capability.
Hard to see the upside for humanity as we've understood it to date, but its definitely going to be more impactful than just making a buck.
I think we literally have those agents today, albeit implemented in meat rather than silicon. Any particular reason you elect not to use the free-to-you travel agent? Generally they are the same or less expensive and able to work in your best interests.
Because a technically-sophisticated enterprise generally wouldn't pick Jamf unless they were an exclusive Apple house - they'd consider something like VMware's Workspace One (now Omnissa) that works across windows, mac, and linux.
And no need to ask an LLM - we can read the doc and notice there are 6 different deployment methods that work (out of the box) for the built-in MacOS MDM, several of which allow transparent, mandatory updates to user environments without requiring user interaction.
It easy to convince a college student with $20 that an ikea chair is good. Artisanal is overkill for plenty of scenarios, and definitely those where time or money are constrained.
"Consent" and "Legitimate Interest" are legal terminology - they're two bases defined in GDPR and have different implications and requirements for balancing user and processor interests.
When the author says that Persona claims the "legitimate interest" basis for these data, they're saying that Persona is trying to achieve maximum flexibility for using the data (since "consent" generally requires specific agreement on a specific use for the data, and the burden of maintaining the consent records, where "legitimate interest" does not).
Sometimes, but if your product is a platform (or anything beyond a niche solution in a broader problem domain) then you're going to spend some time showing people how to make it work best for them.
This chart shows total tax receipts as a percent of GDP, which doesn't seem to address the poster's contention that historically the rich paid a higher share of those receipts through elevated marginal tax rates.
They don't NEED them, but better project estimates can reduce the error bars on other dependent estimates (e.g. estimated sales, estimated ship dates, estimated staffing requirements, etc...), and that might be useful to a business (or not).
Plenty of my former coworkers have evolved into lifelong substantial friendships.
What started with smalltalk evolved into conversations over lunch which then afforded after work socializing which then led to actively scheduling time for shared interests. All of those provided ample opportunity to learn almost everything about that person and open the door to a deep friendship when mutually desired.
The code was just created to support some broader goal, which it presumably did much of the time. The value of those goals is where the meaning comes from.
A chef reflecting on their life would hardly lament that every meal they'd ever crafted ended up in the bin (or the toilet).
Much of the coding we do is repetitive and exists in the training data, so I think its pretty great if AI can eliminate that toil and liberate the meat to focus on the creative work.
LLMs can contribute quite reliably given very narrow prompts and short horizons (keeping turns low and context brief). If you chain a bunch of these narrow contributions together and define guardrails (structured outputs, online evals, other-llm-as-judge/jury, etc...) you can produce a very repeatable workflow that reliably delivers to defined service levels.
The obvious issue being - you've got to define the workflow and implement all the guardrails, not hope that the LLM will infer them during a session or a one-shot prompt.