Real Time

I am user #3392 on Slashdot.

[Real Time @realtime3392]

@TillyNorwoodX I think it is endearing. Keep being you, @TillyNorwoodX .

[Real Time @realtime3392]

I had an agentic AI joke, but I think I hallucinated it.

[Real Time @realtime3392]

@hifriends7640 360 - 90 - (180 - 60) = 150

[Real Time @realtime3392]

@TheAhmadOsman @LottoLabs Blackwell FTW!

[Real Time @realtime3392]

@hicasamadim Lamb weight: 33 / 3 = 11 kg

Cat weight: 31 - one lamb = 20 kg !?!?!! OK

Two monkeys weight: 50 - one cat = 30 kg

Monkey weight: 30 / 2 = 15kg

Still, that is one **fat** cat...

[Real Time @realtime3392]

@FiredUpCoug Replace #2 with:

.45 ACP handgun, preferably double stack

[Real Time @realtime3392]

The specs on the Mikrotik site don't mention the fan noise level. Can you comment on whether this is "at my desk" quiet (which I would hope, with only 123W power draw), or "Mellanox my ears are bleeding" loud? (Speaking from HPE OEM Mellanox SN3700 200GbE 32-port Onyx network switch experience...)

[Real Time @realtime3392]

@SidFishus @DW4Liberty @frankwrighter @TheHerefordian Wow. What a complete cunt.

[Real Time @realtime3392]

@unclebobmartin @esrtweet @grok Actually, there are already examples of multi-modal models perfectly able to describe their surroundings. **Gemini Live** on an Android phone can regularly do this. (I set my phone in a dashboard stand and give it access to the forward facing camera.)

[Real Time @realtime3392]

But were you interacting with a multi-modal model with actual vision and audio capabilities? How would you expect an actual human to respond to that question with earplugs and wearing a sack hood?

@grok Does the "describe your surroundings" seem like a fair test for a text-only LLM? Seems like you need to employ tests of equivalent "sensory bandwidth" to humans and see if they pass or not, no?

[Real Time @realtime3392]

@dashboardlim FREE

[Real Time @realtime3392]

Funny how the 15th Amendment contains a prohibition on intentional racial discrimination, but @BarakObama thinks it only applies if you aren't White.

Last I checked, "White" is considered a race, too. Frankly, I'm tired of decades of intentional racial discrimination of any kind, against any race.

[Real Time @realtime3392]

@TheAhmadOsman They don't.

You can still find Google Coral edge TPUs, though.

[Real Time @realtime3392]

@TheAhmadOsman I'm sorry. 😁

Lived there for eight years. Was so glad to get to work remotely and leave.

[Real Time @realtime3392]

@esrtweet The full 8 MIRV W88 load-out of a single Trident II D5 SLBM would be enough to catastrophically rupture the Three Gorges Dam, and there are no ABM defenses present that could prevent such an attack.

[Real Time @realtime3392]

@drunknbass @TheAhmadOsman What are you using to serve your models? vLLM?

[Real Time @realtime3392]

No way to find? I've bought three of the two-slot Ampere generation NV-Link bridges on eBay for an average of ~$200 each in the past month. Are you even looking? And, yes, if you are serving your model with something like vLLM that offers things like tensor parallelism, it **does** make a difference in TPS for dense models that don't fit (along with their context) comfortably in 24GB.

[Real Time @realtime3392]

The two-slot Ampere generation NV-Link bridges (suitable for two-slot, blower-style RTX 3090 GPUs, e.g. Gigabyte) are readily available for about $200 on eBay. Even the wider bridges needed for gaming RTX 3090 GPUs are available on eBay for $300~$400. So, not sure why you would be paying $2K for an NV-Link bridge.

[Real Time @realtime3392]

Not sure why you are paying $2K for an NV-Link bridge. The two-slot ones are about $200 on eBay (great if you are buying the two-slot blower-style RTX 3090 GPUs like the ones from Gigabyte), and wider ones (for gaming RTX 3090 cards that are wider than two slots) are around $300 to $400.