You can entirely remove the word “forensics” and my statement would be the same…
4,694 karma · joined August 16, 2023
AI images of real life still don’t pass intense scrutiny.
I don’t think this would be immune to that.
Yea, duh.
The more your decisions affect the company, the more you’re compensated.
It’s why C-suite commands the biggest salaries.
It’s another reason why orgs are so low trust.
People’s unawareness of their own personification bias with LLMs is wild.
My question was what do cars let us do now that we couldn’t before.
We moved heavy objects and traveled long distances long before cars.
They make it easier and faster, but don’t add new capabilities
I don’t think most republicans like trump, he just has a cult following.
I also don’t think that any other republican can garner that kind of following. Trump has aura in a way that no other public figure has right now.
I don’t think any regime he sets up would survive his death.
It’s as bad as untangling the last guy’s typescript spaghetti. He quit, so I can’t ask him about it either.
… devs don’t hire the teams… management does…
Most devs I worked with prefer smaller teams.
A little pessimistic, no?
People thought the blue wave would never end after Obama was elected.
No way to tell what the future holds.
If your leadership is good and competent, they can drive the company any way they feel and have it work.
The most productive I’ve been in my career was when my PO put all tasks that needed to be done in a Google sheet and the whole dev team spent a week just picking tasks off the list. It ended with us shipping our app on time.
No planning, no grooming, no nothing. It was a high trust, high ownership team.
I miss those days.
Ofc decisions can have consequences, but that’s what the high pay should be for.
It’s not a black and white issue and I don’t think you need to present it like one.
Oh jeez, then we may have just been talking past each other. I thought that’s what you were arguing for.
> That just isn't important here at all.
It is, though. The fact that the underlying processes are well understood means that, if we so wished, we could work backwards and understand what the model is doing.
I recall some papers on this, but can’t seem to find them right now. One suggested that groups of weights relate to specific kinds of high level info (like people) which I thought was neat.
> the comparison wasn't intended they wouldn't be called "artificial intelligence"
Remember “smart” appliances? Were we meant to compare an internet connected washing machine to smart people? Names are all made up.
I do actually think AI is a horrible name as it invites these kinds of comparisons and obfuscates more useful questions.
Machine Learning is a better name, imo, but I’m not a fan of personifying machines in science.
Too many people get sci-fi brain.
It’s just the easiest way to make the stock price go up.
Layoffs caused twitter to bump in price. Same with Microsoft when they laid off all those game devs.
C-suite don’t care about profit or losses in tech, just the stock price.
In the 2010s, hiring devs increased stock prices. In the 2020s mass firings do.
It’s just money. Simple as.
It’s not the devs who are bloating development costs. It’s the layers of management making hour long meetings to discuss button placement.
It’s the hours of retros, design meetings, and skip-levels designed to remove any personal investment or sense of ownership from everything.
Since so few individuals are trusted with real decision making power, you need a lot more people to achieve some kind of consensus/buy-in so that you can ship anything.
The devs are at the extreme bottom of this totem poll.
It’s like blaming construction workers for houses being expensive.
And, ofc, that goes without mentioning the multimillion yearly bonuses for C-suite, but we can just forget about that and blame the lowest ranking corporate employees (devs)
Source: I worked at a large tech firm for many years and saw this over and over again.
You can’t prove intention, but I can show examples of LLMs lacking intent (as when repeating the same solution even after being told it was incorrect)
> Generating text is the trace of an internal process in an LLM.
Not really sure precisely what you mean by trace, but the output from an LLM (as with any statistical model) is the result of the calculations, not a representation of some emergent internal state.
If the improvement curve is linear, it would be prohibitively expensive as and if logarithmic, it’d be impossible.
There are 2 things we’re talking about here.
There’s the physical, mechanical operations going on during inference and there’s potentially a higher order process happening as an emergent property of those mechanical operations.
We know precisely the mechanical operations that take place during inference as they are machine instructions which are both man-made and very well understood. I hope we can agree here.
Then there’s potentially a higher order process. The existence of that process and what that process is still a mystery.
We do not know how the human brain works, physically. We can’t inspect discrete units of brain operations as we can with machine instructions.
For that reason, it is uncritical to assume that there is any kind of “thought” process occurring at inference which is similar to our thought processes.
Comparing the two is like apples and oranges anyway and is pedantic in a non-useful way, especially with our limited understanding of the human brain.
I’m not sure any of that was relevant to what I was saying.
And you still didn’t answer my question.
AI models only improve through training and good luck convincing any given LLM provider to improve their models for your specific use case unless you have deep pockets…
Yes and no.
Lots of quick sweeping local changes were promised to specific states during trump’s rallies in those states only for him to go silent on them post election.
I don’t think flip flopping on tariffs was part of his platform either.
But generally, yes, this is what was voted for.
He takes the gish gallop approach to governing, so it’s hard to make any large statements like this without being a little incorrect.
Also why bring up disliking something as though you were ahead of the curve only to stop short of actually being in favor of taking action?
Goodheart’s law states: “ When a measure becomes a target, it ceases to be a good measure”
If your accounting is too fine grained, there becomes an inescapable draw to point at a metric and declare that a goal.
Once that happens, engineering tasks can be come extremely well defined to the point of acting like tunnel vision.
Instead of finding improvements or innovations generally, the focus is entirely on moving a specific metric.
You see this issue in highly metric driven orgs as well.
How could we not know? Every processor instruction is observable.
What we specifically don’t have a good view is the causal relationship between input tokens, a model’s weights, and the output.
We don’t know specifically what weights matter or why.
That’s very different than not understanding what processes are taking place.