Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > sci.physics.relativity > #667219

The joy of running an AI Model on your AI Laptop [Local AI] (Was: OpenAI is Toast: Kimi K2 is Live [Halting Problem])

Cross-posted to 3 groups.

Show all headers | View raw

Hi,

Now that was a little unexpected that Intel beats AMD.
Could get my hands on two AI Laptops for testing.
The have both DDR5X Ram.

AI Laptop 1: Acer_Swift_SFG16-61
AMD Ryzen 7 350
Radeon Graphics
AMD Ryzen AI Boost, up to 50 TOPS

AI Laptop 2: LENOVO_83JQ
Intel Core Ultra 7 258V
Intel Arc Graphics 140V
Intel AI Boost, up to 47 TOPS

I couldn't test the AI Boost yet. Besides the nice
CPUs, the iGPU and the fast memory, in this case
32 GB, seem to give quite some bang:

Here some test with LMStudio, I picked the
model Qwen3 Coder 30B, with Q4_K_M encoding,
which is around 16 GB, and fits into my GPU memory:
https://lmstudio.ai/models/qwen/qwen3-coder-30b

AI Laptop 1: 18 Tokens/s

AI Laptop 2: 6 Tokens/s

Now I am waiting for drivers and applications
that make use of the AI Boost. Will say
I only tested the iGPU, not yet the NPU.

For the donwload of the AI model it is
required to have fiber and WiFi 6, you can
then make it in less than a minute, even

if you are far away from the router.

Bye

P.S.: The 18 Tokens/s are quite fluid,
the 6 Tokens/s are a little annoying. But
on both machines I could converse with the

model, and go from primes to twin primes.
It didn't use any RAG, fully produce via
Local AI, also remote AI access after

download of the AI Model:

----------------- cut here ---------------------

Here's a Prolog implementation to find
twin prime pairs (pairs of primes that differ by 2):

% Generate twin prime pairs up to N
twin_primes_up_to(N, TwinPairs) :-
     findall((P, Q),
             (between(3, N, P),
              is_prime(P),
              Q is P + 2,
              is_prime(Q)),
             TwinPairs).

% Check if a number is prime
is_prime(2) :- !.
is_prime(3) :- !.
is_prime(N) :-
     N > 3,
     N mod 2 =\= 0,          % Even numbers > 2 are not prime
     \+ has_factor(N, 3).    % Check if N has any factors starting from 3

% Check if N has a factor starting from F
has_factor(N, F) :-
     F * F =< N,
     (N mod F =:= 0 ; has_factor(N, F + 2)).

Example usage:

?- twin_primes_up_to(20, Pairs).
Pairs = [(3, 5), (5, 7), (11, 13), (17, 19)].

?- twin_primes_up_to(50, Pairs).
Pairs = [(3, 5), (5, 7), (11, 13), (17, 19), (29, 31), (41, 43)].

This finds pairs of primes (P, Q) where Q = P + 2,
which are called twin primes. The findall/3 predicate
collects all such pairs where both P and P+2 are
prime numbers and P is less than or equal to N.

----------------- cut here ---------------------

Mild Shock schrieb:
> Hi,
> 
> Lets analyse Ross Finlaysons Halting Problem
> nonsensen with Kimi K2:
> 
> Ross Finlayson schrieb:
>  > I don't go looking for paradoxes,
>  > since there aren't any in a real theory,
>  > contradictions though automatically
>  > make for counterexamples, then about
>  > the "analytical bridges",
>  > or ponts is what I call them,
>  > not the "invincible ignorance".
> 
> Lets see what Kimi K2 says:
> 
> Q: In the arithmetic hierarchy counter models are
> a level higher, than provability.
> 
> A: I think you're referring to a well-known fact
> in logic: the set of (codes of) sentences that
> are true in some model of arithmetic is one level
> higher in the arithmetical hierarchy than the set
> of sentences that are provable in, say, Peano Arithmetic.
> 
> Let me double-check the precise levels and any subtleties.
> Let me bring in a classic reference that spells out the
> exact levels. The key fact is:
> 
> The set of (Gödel numbers of) sentences provable in
> Peano Arithmetic is recursively enumerable, hence Σ₁⁰.
> The set of sentences that are true in some model of
> PA is Σ₂⁰ (and complete for that level).
> 
> A concise statement of this can be found in Hinman’s
> textbook or in the classic survey by Kreisel & Levy.
> Let me look for an explicit citation.
> 
> Exactly — the blog post I just fetched already says
> it in plain language:
> “∃y (y Gödel-codes a proof from PA of the sentence Gödel-coded by x) … 
> defines the set of natural numbers that encode theorems of PA. … This 
> set is … Σ₁.”
> So the set of provable sentences is Σ₁.
> A couple of paragraphs later it continues:
> “Σ₂ sets are those that can be defined by a sentence that
> has a block of ∃s, then a block of ∀s, and finally
> a Σ₀ sentence …
> …
> ∃y ∀z (TM(x) doesn’t halt on input y after z steps)
> … defines the set of Turing machines that compute partial functions.”
> That set is Σ₂ and is not Σ₁.
> 
> Now let me connect the dots explicitly:
> Provability = “there exists a proof” ⇒ one
> unbounded ∃ ⇒ Σ₁.
> Satisfiability (having a model) = “there exists a
> model” followed by “for every axiom, it is true in
> that model” ⇒ ∃ followed by ∀ ⇒ Σ₂.
> Hence, as your couplet says, In the arithmetic hierarchy,
> counter-models are a level higher than provability.
> 
> they indeed sit one step higher: Σ₂ vs. Σ₁.
> 
> Click the link to view conversation with Kimi AI Assistant 
> https://www.kimi.com/share/19a6e410-3582-81e2-8000-00000e33bfe3
> 
> Bye
> 
> Mild Shock schrieb:
>> Hi,
>>
>> So the Tibetan lamasery had it right,
>> the lights go out. There is a never ending
>> hunger for crunch? Even requiring USA
>>
>> to go into domestic chip production:
>>
>> America’s Most Advanced Chip Factory Yet
>> https://www.youtube.com/watch?v=1VX3jNJmbcI
>>
>> But chip production etc.. (embodied energy) is
>> only a fraction of the energy consumption:
>>
>>  > Operational energy (running the data center
>>  > for 1 year at 1 GW) = 8,760 TWh Hardware
>>  > manufacturing energy = 0.75 TWh
>>  > So the energy to produce the hardware is
>>  > roughly 0.0085 (or 0.85%) of the annual
>>  > operational energy.
>>
>> What are the projections? By 2030, global
>> power demand for data centers is projected to
>> reach approximately 220 GW, underscoring the
>>
>> urgency and strategic importance of securing
>> power and hardware capacity:
>>
>> $100B Bet on 10GW AI Infrastructure
>> https://51ai.substack.com/p/openai-nvidia-100b-bet-on-10gw-ai
>>
>> Bye
>>
>> Mild Shock schrieb:
>>> Hi,
>>>
>>> Interestingly there is also now a top-ten
>>> for AI data centers, not only super computers.
>>> We are talking about newly built AI data centers
>>>
>>> that for the first time go into giga watts:
>>>
>>> The New World’s Largest AI Supercluster
>>> https://www.youtube.com/watch?v=RxuSvyOwVCI
>>>
>>> Some confirmed picks:
>>>
>>> xAI Colossus Memphis
>>> Phase 1    100000    xAI    U.S.    Confirmed
>>> Tesla Cortex Phase 1    50000    Tesla    U.S.    Confirmed
>>> Lawrence Livermore NL
>>> El Capitan Phase 2    44143    U.S. Department of
>>> Energy    U.S.    Confirmed
>>> Anonymized Chinese
>>> System    30000    N/A    China    Confirmed
>>> Meta GenAI 2024a    24576    Meta AI    U.S.    Confirmed
>>> Meta GenAI 2024b    24576    Meta AI    U.S.    Confirmed
>>> Jupiter, Jülich    23536    EuroHPC JU,
>>> Jülich Supercomputing
>>> Center    Germany    Confirmed
>>> Anonymized Chinese
>>> System    20000    N/A    China    Confirmed
>>> Anonymized Chinese
>>> System    20000    N/A    China    Confirmed
>>>
>>> https://www.visualcapitalist.com/the-worlds-most-powerful-ai-supercomputers/ 
>>>
>>> Bye
>>>
>>> Mild Shock schrieb:
>>>>
>>>>
>>>> In a Tibetan lamasery, the monks seek to
>>>> list all of the names of God. They believe
>>>> the Universe was created for this purpose,
>>>> and that once this naming is completed, God
>>>> will bring the Universe to an end. Three
>>>>
>>>> centuries ago, the monks created an alphabet
>>>> in which they calculated they could encode
>>>> all the possible names of God, numbering
>>>> about 9,000,000,000 ("nine billion") and
>>>> each having no more than nine characters.
>>>>
>>>> riting the names out by hand, as they had
>>>> been doing, even after eliminating various
>>>> nonsense combinations, would take another
>>>> 15,000 years; the monks wish to use modern
>>>> technology to finish this task in 100 days.
>>>>
>>>> They rent a computer capable of printing all
>>>> the possible permutations, and hire two
>>>> Westerners to install and program the machine.
>>>> The computer operators are skeptical but
>>>> play along. After three months, as the job
>>>>
>>>> nears completion, they fear that the monks
>>>> will blame the computer (and, by extension,
>>>> its operators) when nothing happens. The
>>>> Westerners leave slightly earlier than their
>>>> scheduled departure without warning the monks,
>>>>
>>>> so that it will complete its final print run
>>>> shortly after they leave. On their way to the
>>>> airfield they pause on the mountain path. Under
>>>> a clear night sky they es timate that it must be
>>>> just about the time that the monks are pasting
>>>>
>>>> the final printed names into their holy books.
>>>> Then they notice that "overhead, without any
>>>> fuss, the stars were going out."
>>>>
>>>> https://en.wikipedia.org/wiki/The_Nine_Billion_Names_of_God
>>>
>>
> 

Back to sci.physics.relativity | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

OpenAI is Toast: Kimi K2 is Live [Halting Problem] (Was: USA domestic chips for the never ending crunch) Mild Shock <janburse@fastmail.fm> - 2025-11-10 15:57 +0100
  Re: OpenAI is Toast: Kimi K2 is Live [Halting Problem] (Was: USA domestic chips for the never ending crunch) olcott <polcott333@gmail.com> - 2025-11-10 12:20 -0600
    Re: OpenAI is Toast: Kimi K2 is Live [Halting Problem] (Was: USA domestic chips for the never ending crunch) "Chris M. Thomasson" <chris.m.thomasson.1@gmail.com> - 2025-11-10 12:08 -0800
  Re: OpenAI is Toast: Kimi K2 is Live [Halting Problem] (Was: USA domestic chips for the never ending crunch) Ross Finlayson <ross.a.finlayson@gmail.com> - 2025-11-10 17:05 -0800
  The joy of running an AI Model on your AI Laptop [Local AI] (Was: OpenAI is Toast: Kimi K2 is Live [Halting Problem]) Mild Shock <janburse@fastmail.fm> - 2025-11-11 21:09 +0100
  The joy of running an AI Model on your AI Laptop [Local AI] (Was: OpenAI is Toast: Kimi K2 is Live [Halting Problem]) Mild Shock <janburse@fastmail.fm> - 2025-11-11 21:12 +0100
    Cute Racoon generated with Local AI (Was: The joy of running an AI Model on your AI Laptop [Local AI]) Mild Shock <janburse@fastmail.fm> - 2025-11-11 23:26 +0100
  Meta is Dust: Yann LeCun crushed by Behemoth China (Was: OpenAI is Toast: Kimi K2 is Live [Halting Problem]) Mild Shock <janburse@fastmail.fm> - 2025-11-13 19:03 +0100
    Even Prolog now depends on China (Was: Meta is Dust: Yann LeCun crushed by Behemoth China) Mild Shock <janburse@fastmail.fm> - 2025-11-13 19:13 +0100
      Re: Even Prolog now depends on China (Was: Meta is Dust: Yann LeCun crushed by Behemoth China) Ross Finlayson <ross.a.finlayson@gmail.com> - 2025-11-13 11:42 -0800

csiph-web