6 ms·
I got one a couple of days ago, and it really threw me for a loop. I'm used to ChatGPT at least being coherent, even if it isn't always right. Then I got this a
by nrclark 3y ago
I got one a couple of days ago, and it really threw me for a loop. I'm used to ChatGPT at least being coherent, even if it isn't always right. Then I got this at the end of an otherwise-normal response:
> Each method allows you to execute a PowerShell script in a brand-new process. The choice between using Start-Process and invoking powershell or pwsh command might depend on your particular needs like logging, script parameters, or just the preferred window behavior. Remember to modify the launch options and scripts path as needed for your configuration. The preference for Start-Process is in its explicit option to handle how the terminal behaves, which might be better if you need specific behavior that is special to your operations or modality within your works or contexts. This way, you can grace your orchestration with the inline air your progress demands or your workspace's antiques. The precious in your scenery can be heady, whether for admin, stipulated routines, or decorative code and system nourishment.
- engineer_22 3y agoIt reads like a bad Chinese translation :)
- runeofdoom 3y agoThat last part sounds like the Orz from Star Control II. Almost sensical, in a vaguely creepy way. Like an uncanny valley for langauge.
- namaria 3y agoRealizing that the model isn't having a cogent conversation with the user, that the output unravels into incoherence as you extend it enough and that the whole shock value of ChatGPT was due to offering a limited window where it was capable of sorta making sense was the realization that convinced me this whole gen ai thing hinges way more on data compression than simulated cognition of any sort.
- golergka 3y agoWhy do you think that data compression and cognition are fundamentally different?
- namaria 3y agoThe behavior of large language models compressing 20 years of internet and being incapable of showing any true understanding of the things described therein.
- FredPret 3y agoA human also compresses many years of experience into one conversation. Does this reflect true understanding of the things described? Only the human doing the talking can know, and even that is on shaky ground. (if you don't understand something, will you always realize this? You have to know it a little bit to judge your own competence).
- namaria 3y agoWe compress data from many senses and can use that to interactively build inner models and filters for the data stream. The experience of psychedelics such as psylocibin and lsd can be summarized as disabling some of these filters. The deep dream trick google did a while back was a good illustration of hallucinations and also seen in some symptoms of schizophrenia. In my view that shows we are simulating some brain data processing functions. Results from the systems conducting these simulations are very far from the capabilities of humans but help shed light into how we work. Conflating these systems with the full cognitive range of human understanding is disingenuous at best.
- resource0x 3y agoMy theory is that the system ate one terabyte too many and couldn't swallow. Too much data in the training set might not be beneficial. It's not just diminishing returns, but rather negative returns.
- kordlessagain 3y agoGPT-3.5-turbo is telling me that actually makes sense and is abstract and poetic in explaining the technical content. > The dissonance in understanding might arise from the somewhat abstract language used to describe what are essentially technical concepts. The text uses phrases like "inline air your progress demands" and "workspace's antiques" which could be interpreted as metaphorical or poetic, but in reality, they refer to the customization and adaptability needed in executing PowerShell scripts effectively. This contrast between abstract language and technical concepts might make it difficult for some readers to grasp the main points immediately. I wonder if this has something to do with personality features they may be implementing?
- jimmux 3y agoI think that's more due to GPT's need to please, so if you ask it to make sense of something it will assume there is some underlying sense to it, rather than say it's unparsable gibberish.
- kordlessagain 3y agoI had it do it with several anecdotical reports, and it said those were nonsense, where this one it said made sense and explained why. Metaphorically speaking is a thing, and it doesn't make it inaccurate, just a bit odd.
- irthomasthomas 3y agoThe strangest thing about this issue, the meltdown happened on every model I tried: 3.5-turbo, 4-turbo, and 4-vision where all acting dumb as dirt. How can this be? There must be a common model shared between them, a router model perhaps. Or someone swapped out every model with a 2bit quantized version?