gmueckl

7 ms·

121.
by gmueckl 4mo ago
A lot pf good products are a combination of features that customers need and use and features they think they need and ask for, but never use. But the sales wouldn't be as good without them. It's a bit comical once it becomes appa
122.
by gmueckl 4mo ago
The first time this happens in a petri dish will likely have to be under extremely controlled circumstances. But the process will be modified and toyed with once it exists and I think that this will eventually lead to whole spectrum of (qua
123.
by gmueckl 4mo ago
I have not enough imagination to come up with a scenario where making a sandwich would create concern that it might obstruct an investigation. But obstruction is defined by intent and outcome, not the exact means. Abstractions like this are
124.
by gmueckl 4mo ago
It is a long time and it gets frustrating when there is significant time where there is flailing with no visible progress. I have had long bug hunts (~a month each) and witnessed ones that took much, much longer. But the longest one I witne
125.
by gmueckl 4mo ago
Humans are not flawless, but they are much, much better at reasoning than LLMs are. LLMs can be made to fail quite reliably and easily because they cannot build proper manipulatable/predictive models. This is related to the point that
126.
by gmueckl 4mo ago
This is exactly backwards. The brittleness is because they emulate reasoning without actually algorithmically performing it. Add.: I pointed to this class of problems specifically because they require the ability to abstract in a way that t
127.
by gmueckl 4mo ago
Reasoning includes things like proper use of logic. LLMs have been repeatedly shown to fail horribly at this. They consistently fail at drawing basic logical conclusions because they cannot build a sufficiently abstract model of certain pro
128.
by gmueckl 4mo ago
What else is there to say? LLMs can at most regurgitate approximations of human reasoning steps in the limited forms in which they may be expressed in the training data or interpolations thereof. That's the core essence of what they ar
129.
by gmueckl 4mo ago
LLM output produces the illusion of reasoning. The underlying computation, however, is not reasoning.
130.
by gmueckl 4mo ago
Not a valid comparison. Chess algorithms are built around the rules of chess, most notably the turn taking nature of the game (min/max with alpha/beta pruning based on lists of valid moves in any position).
131.
by gmueckl 4mo ago
There cannot be any reasoning embedded in the model. The algorithm is literally "predict the most likely next token". Anything beyond that is just patterns in the predictions fooling us humans into ascribing more to the system tha
132.
by gmueckl 4mo ago
The reason people cite when defending Github usually isn't the repisitory hosting itself, but the extra features like oull request management and issue tracker that also generate massive network effects by enabling the same accounts to
133.
by gmueckl 4mo ago
Cars are largely not exported to the US from Germany, but built in US factories. German car companies have factories all over the world.
134.
by gmueckl 4mo ago
But money counterfeiting is a different proposition: the printer is blocking exactly known patterns found on real currency. In fact, many bill designs incorporate patterns that are easily machine detectable for this purpose. Blocking the pr
135.
by gmueckl 4mo ago
I feel like there is a lot pf potential for display-based AR in a glasses form factor. I wouldn't want to dismiss that outright. However, the concern about sensors (cameras and microphones) in the glasses is also real, despite them add
136.
by gmueckl 4mo ago
Seeing PlasticSCM wither like this is sad. I had rather high hopes for this system to successfully compete with Perforce, but history worked out differently.
137.
by gmueckl 4mo ago
Tools that support non-destructive editing workflows could in theory provide limited diff and merge capabilities on their internal graphs pf non-destructive operations. E.g. Photoshop could in theory merge two files where unrelated layers h
138.
by gmueckl 4mo ago
Merge conflicts on binary files are typically unresolvable in practice because the editing tools don't normally have the ability to compare files. This leads to one of the two divergent sets of changes getting lost. That's why gre
139.
by gmueckl 4mo ago
This is ... charitable. Enforcing laws relies on courts which are only reacting slowly to a non-cooperarive administration, which gleefully ignores orders or practices malicious compliance after the damage has been done (and done as fast as
140.
by gmueckl 4mo ago
Executive orders can effectively kneecap any efforts to put laws into effect, as we can see almost daily since 20. January 2025.
141.
by gmueckl 4mo ago
For the sake of this discussion, I'm going with the nationalistic vibe of the order: anybody who isn't a citizen of the USA (presumably to limit risk of AI-supported action against the US?). But that in itself is telling in a way:
142.
by gmueckl 4mo ago
How many of those methods can realistically exfiltrate 20Tb of data? That's quite hard even for well funded actors.
143.
by gmueckl 4mo ago
It depends on the end goal. Free good enough models are a way to drastically devalue Anthropic and OpenAI. A well timed release of a capable model that can run on obtainable hardware, so that a small/medium company can afford self host
144.
by gmueckl 4mo ago
This honestly sounds like a tweaked system prompt more than anything. Maybe it is an attempt to make the model appear stronger?
145.
by gmueckl 4mo ago
It may not be perfect, but this hurdle would still keep out ~99% of the targeted people.
146.
by gmueckl 4mo ago
I would agree if it wasn't for the fact that extracting that volume of data from a properly secured corporate network should be hard. It should raise some flags if a such a high volume of data is downloaded to a user's local machi
147.
by gmueckl 4mo ago
Do you have sources? I would like to read more about that.
148.
by gmueckl 4mo ago
I'm not advocating committing secrets to version control. But you can learn to phrase your WTFs about your colleague's code politely and constructively. I would even argue that this an absolutely basic skill for professionals. I
149.
by gmueckl 4mo ago
Don't be afraid to show your thoughts when asked to. The best developers are those that can express their thoughts clearly at any stage throughout their process. This is one of the skills that shows to me the level of experience a deve
150.
by gmueckl 4mo ago
To give the users a choice? It's not Apple's responsibility to protect the users if they want to opt out.

More ›