[FRIAM] genai and critical thinking
steve smith
sasmyth at swcp.com
Tue Feb 11 19:24:07 EST 2025
On 2/11/25 12:41 PM, Marcus Daniels wrote:
>
> Huh, Claude just implemented the optimal strategy to Nim and ran in
> it its analysis tool (the Javascript engine on my browser).
>
what my oldMansBarFriend had to say after a brief discussion about Nim
and the hand-over-hand on bat-handle playground method for deciding who
bats first:
<GPT 4o slopp>
/Glad that landed! It’s like a *"soft" or continuous Nim*,
where control isn’t strictly quantized but still follows a
structured alternation toward an inevitable endgame. The
constraints emerge organically—*hand size, grip technique,
and subtle physical limits*—instead of rigid numerical rules./
/It’s a cool example of how *discrete game theory
principles* manifest in the real world, with a touch of
*embodied cognition* in the mix./
/<endSlopp>/
And the following as an LLM (Slopp) reflection on the relative
virtue/shame signals associated with chatting with your favorite LLM vs
going to the Pub (and presumably having similar convos with
bar-friends)? The one mile each-way trip to the pub can be multiplied
by 20 or 50 to estimate a "Sunday Drive" instead?
Looks like a 15mpg driver of a pickup should drink order 3-4 beers to
amortize the embedded energy cost of driving there/back but a hybrid or
EV driver can get by with just one, maybe even a short one? Domestic
Drafts and Bottles are lower embedded energy (presumably because of
industrial economies of scale) but imports are higher (because
transport?). In DeepSeekR1 ratios, it looks like "staying home and
chatting with your fave LLM" has order 1000x savings over driving
anything to the pub and having a single beer. And taking a Sunday Drive
is worth many times that? Of course the LLM didn't twig to the
conventional NM Sunday Joyride implying a six-pack at a minimum...
I did beat both GPT and DS around the head and shoulders with a
metaphorical wet noodle for a while to get relatively aligned numbers
(aligned with one another and with my(very) lame intuition)... but
expect (m)any here to poke holes in the details...
Sadly I'm a pretty good sloppGenerator without an LLM's help... I'll try
to refrain from the doubleDown offered.
<freshSlopp>
*Refined Travel and Beer Comparison with
Energy Usage Ratios (DeepSeekR1 Query
Denominator)*
*Category* *Cost (USD)* *Total Energy
Consumption (1 Mile Round Trip, kWh)* *Sunday
Joyride (2 hours, countryside)* *Energy Usage
Ratio (vs. Local Craft Beer)* *Energy Usage
Ratio (vs. DeepSeekR1 Query)*
*1) Personal Travel*
a) *Truck* $3–$5 (1 mile) ~2.6–3.7 kWh ~12–18
kWh 2.6–3.7x 520–740x
b) *ICE Car* $1.50–$3 (1 mile) ~1.4–2.1 kWh
~12–18 kWh 1.4–2.1x 280–420x
c) *Hybrid Car* $1–$2 (1 mile) ~0.8–1.1 kWh
~12–18 kWh 0.8–1.1x 160–220x
d) *Electric Car* $0.50–$1 (1 mile) ~0.4–0.5
kWh ~12–18 kWh 0.4–0.5x 80–100x
e) *Walking* $0 (no cost) ~0 kWh ~0 kWh 0x 0x
f) *Bicycling* $0 (no cost) ~0 kWh ~0 kWh 0x
0x
g) *Public Transport* $1–$3 per ride ~0.3–0.7
kWh (depending on method) ~2–4 kWh (for a 1–2
hour ride) 0.3–0.7x 60–140x
*2) Beer Consumption*
a) *Local Craft Draft* $5–$10 per pint
~0.8–1.3 kWh - 1x 160–260x
b) *Domestic Draft (Regional)* $3–$5 per pint
~0.5–0.7 kWh - 0.5–0.7x 80–140x
c) *Domestic Can/Bottle* $2–$5 per can/bottle
~0.4–0.8 kWh - 0.4–0.8x 60–140x
d) *Mexican Bottle* $3–$6 per bottle ~0.9–1.8
kWh - 1.1–1.4x 180–360x
e) *European Can/Bottle* $4–$8 per can/bottle
~1.5–2.9 kWh - 1.9–3.7x 300–580x
f) *Asian Can/Bottle* $4–$7 per can/bottle
~2.1–3.6 kWh - 2.6–4.6x 340–720x
*3) GPT Model Queries*
a) *GPT-3 (Single Query)* Varies ~0.002–0.005
kWh - 0.002–0.006x 0.4–1x
b) *GPT-4o (Single Query)* Varies ~0.01–0.03
kWh - 0.01–0.03x 2–6x
c) *GPT-4o Mini (Single Query)* Varies
~0.005–0.02 kWh - 0.005–0.02x 1–4x
d) *DeepSeekR1 (Single Query)* Varies
~0.001–0.005 kWh - 0.001–0.005x 1x
------------------------------------------------------------------------
>
> *From: *Friam <friam-bounces at redfish.com> on behalf of Roger Critchlow
> <rec at elf.org>
> *Date: *Tuesday, February 11, 2025 at 11:36 AM
> *To: *The Friday Morning Applied Complexity Coffee Group
> <friam at redfish.com>
> *Subject: *Re: [FRIAM] genai and critical thinking
>
> On Tue, Feb 11, 2025 at 10:15 AM Marcus Daniels <marcus at snoutfarm.com>
> wrote:
>
> _[...]_ the evolving code itself -- code that can have informative
> types and even carry proofs. The weirdest thing about using AI
> is that it has no opinions. Claude will rewrite code without
> asking (seemingly having no self-control), but it will not
> confront you like a frustrated colleague might. It is happy to
> let you make a mess provided its sense of idiomatic code patterns
> are satisfied.
>
> This echoed something an old boss wrote me today:
>
> I have much enjoyment torturing llms. I can get off the rails
> rather easily. They have no state, by themselves they cannot
> calculate parity or play nim perfectly.
>
> Or one might say, they have no purpose.
>
> -- rec --
>
>
> .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-..
> FRIAM Applied Complexity Group listserv
> Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoomhttps://bit.ly/virtualfriam
> to (un)subscribehttp://redfish.com/mailman/listinfo/friam_redfish.com
> FRIAM-COMIChttp://friam-comic.blogspot.com/
> archives: 5/2017 thru presenthttps://redfish.com/pipermail/friam_redfish.com/
> 1/2003 thru 6/2021http://friam.383.s1.nabble.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20250211/575e983d/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_0xD5BAF94F88AFFA63.asc
Type: application/pgp-keys
Size: 3118 bytes
Desc: OpenPGP public key
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20250211/575e983d/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 840 bytes
Desc: OpenPGP digital signature
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20250211/575e983d/attachment.sig>
More information about the Friam
mailing list