You are currently viewing our boards as a guest which gives you limited access to view most discussions and access our other features. By joining our free community you will have access to post topics, communicate privately with other members (PM), respond to polls, upload content and access many other special features. Registration is fast, simple and absolutely free so please, join our community today!
The banners on the left side and below do not show for registered users!
If you have any problems with the registration process or your account login, please contact contact us.
Vancouver Off-Topic / Current EventsThe off-topic forum for Vancouver, funnies, non-auto centered discussions, WORK SAFE. While the rules are more relaxed here, there are still rules. Please refer to sticky thread in this forum.
Deepseek is garbage, I asked it some specific questions about products I know a lot about and instead of saying it doesn't know or trying to learn, it just pulled answers completely out of its ass and gave out blatantly wrong info. What's the use of an AI that confidently gives you paragraphs of wrong information?
It's pretty neat considering the time/money put into it, but it ain't no ChatGPT.
it just pulled answers completely out of its ass and gave out blatantly wrong info. What's the use of an AI that confidently gives you paragraphs of wrong information?
Sounds about the same as most of the other ones then lol.
__________________ 1991 Toyota Celica GTFour RC // 2007 Toyota Rav4 V6 // 2000 Jeep Grand Cherokee
1992 Toyota Celica GT-S ["sold"] \\ 2007 Jeep Grand Cherokee CRD [sold] \\ 2000 Jeep Cherokee [sold] \\ 1997 Honda Prelude [sold] \\ 1992 Jeep YJ [sold/crashed] \\ 1987 Mazda RX-7 [sold] \\ 1987 Toyota Celica GT-S [crushed]
Quote:
Originally Posted by maksimizer
half those dudes are hotter than ,my GF.
Quote:
Originally Posted by RevYouUp
reading this thread is like waiting for goku to charge up a spirit bomb in dragon ball z
Quote:
Originally Posted by Good_KarMa
OH thank god. I thought u had sex with my wife. :cry:
Deepseek is garbage, I asked it some specific questions about products I know a lot about and instead of saying it doesn't know or trying to learn, it just pulled answers completely out of its ass and gave out blatantly wrong info. What's the use of an AI that confidently gives you paragraphs of wrong information?
It's pretty neat considering the time/money put into it, but it ain't no ChatGPT.
As a test, I asked several bots "What type of lubricant is used on keyless works of a modern watch movement?" since I have been studying on that lately.
If it didn't give a specific answer, I asked "What actual product should I use though?"
Gemini gives literally the wrong answer that would cause major problems if someone followed it.
I replied with a query stating that, and it said
Quote:
You are absolutely right to call me out on that. My apologies. Moebius 9010 and 9020 are not suitable for keyless works. They are excellent lubricants, but they are designed for other parts of the movement, such as the gear train. Using them in the keyless works could indeed cause significant problems, including gumming up the mechanism and preventing proper functioning.
And then gave a non standard answer that would still work okay (it recommended an old natural lubricant that would require service sooner than more modern products)
GPT gives one right answer if you ask the second prompt
Claude gives the right answer simply stated. I'll post it here:
Quote:
Modern keyless works typically use Moebius 9504 synthetic oil, which is a low-viscosity lubricant specifically designed for sliding surfaces in watch movements.
Llama gives several results, one which would work but not primary. A couple of them are products I never heard of (but the brands are familiar) and Google gives zero results, so that's weird.
Most answered in a few sentences, some gave generic info on lubricants or what keyless works are. Never more than a paragraph or two
Here is DeepSeek's result:
Quote:
<think>
Okay, so I need to figure out what type of lubricant is used on the keyless works of a modern watch movement. Let me start by breaking this down. First, what exactly are the keyless works? From what I remember, in a watch movement, the keyless works refer to the components that allow the user to wind the watch, set the time, and change other functions like the date without using a key, which was the old method. So, it's the mechanism that includes the crown, stem, winding pinion, sliding pinion, and the various levers and springs that interact with them.
Now, lubrication in watches is crucial because it reduces friction between moving parts, prevents wear, and ensures smooth operation. However, not all parts of a watch movement use the same type of lubricant. Different components might require different viscosities or properties depending on their function, the materials involved, the load they bear, and their exposure to environmental factors like temperature and moisture.
The keyless works involve components that are under relatively low stress compared to, say, the escapement or the gear train, but they still need lubrication to operate smoothly. Since these parts are typically engaged when setting the time or winding the watch, they might not be in constant motion like the gears, but they do need to function reliably when the user interacts with the crown.
I recall that watch lubricants are often synthetic oils or greases, designed to be long-lasting and stable. Some common lubricants used in watches include Moebius oils and greases. For example, Moebius 9010 is a synthetic oil that's used in higher-stress areas, while something like Moebius 9415 is a grease used for parts that need a thicker lubricant. Greases are typically used where components might need a more viscous or adherent substance, perhaps where there's sliding friction or where the lubricant needs to stay in place without migrating.
The keyless works might involve sliding parts, like the setting lever and the clutch wheel, which engage and disengage when switching between winding and setting modes. These sliding actions might benefit from a grease that stays in place and doesn't dry out quickly. On the other hand, pivots or rotational parts might use a lighter oil. However, if the keyless works are mostly levers and sliding surfaces, a grease could be more appropriate.
But wait, I should also consider that over-lubrication can be a problem. Applying too much grease or oil might attract dust or cause the mechanism to gum up over time. So the lubricant used must have the right viscosity and be applied in the correct quantity. Additionally, compatibility with the materials (like brass, steel, or alloys) is important to prevent corrosion or degradation.
I think in many modern watches, especially those from reputable manufacturers, the keyless works might use a specific grease. For instance, Moebius Grease 9415, which is a synthetic grease with a consistency that's suitable for sliding parts. Alternatively, some sources mention using lighter oils for certain parts of the keyless works, but greases are common where there's lateral movement.
Wait, another thought: the stem itself, which is part of the keyless works, might have some rotational parts where it enters the case. The stem might be lubricated with a grease to ensure smooth operation when pulled out or pushed in, and to provide a bit of a seal against moisture. However, the internal components like the sliding pinion and the yoke might use a different lubricant.
I should also check if there's a standard practice or a commonly referenced lubricant for keyless works. From some watchmaking guides or forums, I remember that for the keyless works, a grease like Moebius 941 (which is a lighter grease) or 9504 might be recommended. Alternatively, some people might use Epilam treatment, which isn't a lubricant but a surface treatment to reduce friction and prevent oil from spreading. But that's probably supplementary.
In summary, the keyless works likely use a grease rather than an oil, specifically a synthetic watch grease that's designed for low to moderate stress sliding and rotational movements. The exact type might vary by manufacturer, but common ones include Moebius 941 or 9504. It's important that the lubricant doesn't migrate, so a grease with good adhesion properties is preferable here.
</think>
In modern watch movements, the keyless works—responsible for functions like winding and setting the time—typically utilize a synthetic grease as the primary lubricant. This choice is due to the need for a lubricant that adheres well to sliding and rotational components without migrating, ensuring smooth operation and longevity.
Key Considerations:
Functionality: The keyless works involve sliding parts (e.g., setting levers, clutch wheels) and rotational elements (e.g., stem). Grease is ideal for sliding friction, while light oils might be used sparingly on pivots.
Common Lubricants:
Moebius 9415: A widely used synthetic grease for sliding components, offering stability and minimal migration.
Moebius 9504: A lighter grease suitable for fine mechanisms, often applied to stems and levers.
Material Compatibility: These greases are formulated to avoid corrosion and interact safely with metals like brass and steel.
Application Precision: Over-lubrication is avoided to prevent attracting debris or causing stiffness. Grease is applied sparingly to critical contact points.
In summary, synthetic greases such as Moebius 9415 or 9504 are standard for keyless works, balancing adhesion, durability, and resistance to environmental factors.
(Moebius 9504 is the standard answer btw, so it got there eventually). 9415 is usually used for something else (pallet stones)
Deepseek is garbage, I asked it some specific questions about products I know a lot about and instead of saying it doesn't know or trying to learn, it just pulled answers completely out of its ass and gave out blatantly wrong info. What's the use of an AI that confidently gives you paragraphs of wrong information?
I learned a term of this recently at an Amazon AWS AI event. It's called a "hallucination", similar to how it works for real people.
__________________
|| 18 FK8 | R-18692 | Rallye Red | 6 MT ||
|| SOLD 97 E36 M3 Sedan | Arctic Silver | 5MT ||
|| RIP 02 E46 330ci | Schwartz Black II | 5MT | M-Tech II | Black Cube | Shadowline | Stoff Laser/Anthrazit ||
|| RIP 02 E46 M3 | Carbon Black | 6MT ||
A am learning a bit of python o I asked the local smaller version of deepseek I am running to make me a game. I told it what the game should do. It got on line wrong. I tried multiple times, Ran the code in chatgtp and it fixed that line of code. Deepseek does have a version that's optimized for coding.
The full fat version with 671 billion parameters you need 1342 GB of video ram. That's 16 NVIDIA A100 80GB GPU's. At 22K that's 355k just for the GPU's.
__________________ Until the lions have their own historians, the history of the hunt will always glorify the hunter.
(Moebius 9504 is the standard answer btw, so it got there eventually). 9415 is usually used for something else (pallet stones)
That is not practical though
The <think> portion is what makes deepseek unique to other models, it actually shows the thought process of the AI which allows the users to more effectively ask follow up prompts in order to come up with a better answer.
Most other models hide the thought process behind it, but it will likely be something that more of the big models will start doing because it seems users really find it useful.
__________________
Quote:
Originally Posted by skyxx
Sonick is a genius. I won't go into detail what's so great about his post. But it's damn good!
2010 Toyota Rav4 Limited V6 - Wifey's Daily Driver
2009 BMW 128i - Daily Driver
2007 Toyota Rav4 Sport V6 - Sold
1999 Mazda Miata - Sold
2003 Mazda Protege5 - Sold
1987 BMW 325is - Sold
1990 Mazda Miata - Sold
The <think> portion is what makes deepseek unique to other models, it actually shows the thought process of the AI which allows the users to more effectively ask follow up prompts in order to come up with a better answer.
Most other models hide the thought process behind it, but it will likely be something that more of the big models will start doing because it seems users really find it useful.
I guess if you ignore all the "thinking" it shows, it looks about the same as the others
I guess if you ignore all the "thinking" it shows, it looks about the same as the others
What makes it different is the cost. The cost for a million tokens is a lot less vs chatgtp. You can also just run it locally and pay nothing after the hardware costs. This is big saving for companies using AI.
__________________ Until the lions have their own historians, the history of the hunt will always glorify the hunter.
Considering the shit show that the Twitter acquisition has become(I'm talking about the drop in value and loss in business, not the actual operation of it) with the banks having zero hope of recouping the money lent out, who's gonna lend him 100b to buy out ChatGPT?
Considering the shit show that the Twitter acquisition has become(I'm talking about the drop in value and loss in business, not the actual operation of it) with the banks having zero hope of recouping the money lent out, who's gonna lend him 100b to buy out ChatGPT?
Good point. Never thought about it this way; Elon didn't use his own money, but how does the bank get his money back if Elon doesn't pay. Bank get shares? So if it was lent out Elon still has to pay interest... so not like Elon took the money and ran.
hilarious. Altman responded the social platform X, which is owned by Musk: “no thank you but we will buy Twitter for $9.74 billion if you want.”
Good point. Never thought about it this way; Elon didn't use his own money, but how does the bank get his money back if Elon doesn't pay. Bank get shares? So if it was lent out Elon still has to pay interest... so not like Elon took the money and ran.
hilarious. Altman responded the social platform X, which is owned by Musk: “no thank you but we will buy Twitter for $9.74 billion if you want.”
It's also a generous offer, fidelity has X at $9.4 billion and declining every day.
I was watching a Jamie Dimon interview a few weeks ago, and he was asked about X and Musk. He said Musk could raise billions with a snap of his fingers. He also claimed Musk is our generation’s Einstein. There are a lot of billionaires who would throw money at him.
I was watching a Jamie Dimon interview a few weeks ago, and he was asked about X and Musk. He said Musk could raise billions with a snap of his fingers. He also claimed Musk is our generation’s Einstein. There are a lot of billionaires who would throw money at him.
My problem with this is, Musk doesn't actually do any of the engineer work. Einstein actually knows the math and science behind it.
You are currently viewing our boards as a guest which gives you limited access to view most discussions and access our other features. By joining our free community you will have access to post topics, communicate privately with other members (PM), respond to polls, upload content and access many other special features. Registration is fast, simple and absolutely free so please, join our community today!
The banners on the left side and below do not show for registered users!
