fvv

This is a problem for the employer and for the selection and remuneration of staff (at all levels of the chain and not just those in contact with the customer)

I have the opposite experience with codex 5.3 I had to use 5.2 to design and 5.3-codex to execute , while 5.4 was a better in both, and 5.5 ( all used xhigh) is even better

Just because it's not important to pay attention to CVEs, why not waste the readers' time by creating "fictional" CVEs without a disclaimer in the first line? Just because it's not already difficult to scrape through the information and noise on this internet... especially if it appears on the front page of hackernews

There's one thing "the hackers" haven't considered, though! It's illegal to hack an app in the EU,

so the problem of bypassing age verification by hacking saved files doesn't arise at all!

/s

with those $2k you can have 2xB70, with 1.2Tb/sec and 64G Vram, on linux ( and you can scale further while mac prices increase are not linear 0

I believe that as soon as nuclear fusion becomes operational (and perhaps AI could be of great help in this regard in the next 5 years),

all the carbon storage methods (efficient and otherwise) studied so far could be immediately put into action.

In addition to the immediate reduction in the use of fossil fuels for energy production, the scenario could be completely changed in 10 years, water could be desalinated, desertification reversed, etc.

Free and unlimited energy would be the solution to everything. The question is whether we will get there before it's too late... and perhaps AI is the answer?

20 req/hour is 1 req every 3 min.. you have to think a bit and then write the requests..

1. it's 1.5x , it's quite fast for the level of thinking it has

2. no if you are on subscription, it's the same, at 20$ codex 5.4 xhigh provide way more than 20$ opus thinking ( this one instead really can burn 33% with 1 request, try to compare then on same tasks ) also 8x .. ??? if you need 1M token for a special tasks doesn't hit /fast and vice-versa , the higher price doesn't apply on subscription too..

3. false, i'm on pro , so 10x the base , always on /fast (no 1M), and often 2 parallel instances working.. hardly can use 2% (=20% of 5h limit , in 1h of work ( about 15/20 req/hour) ) , claude is way worse on that imo

imo , the main feature is /fast ... who use 1M context and for what? the model become dumber already at 200K.. it's better to manage the context , and since 5.3, codex is very good at managing it

I got this : https://www.svgviewer.dev/s/DNq3oGMB

pro with the 20$ subscription , from the gemini.google.com site , initially the request was routed to nano banana,

so I had to change the prompt and had to be very specific about not generating the image directly, here the exact prompt :

"I want the html svg code for that ( not the generated image) , generate an SVG of a pelican riding a bicycle"

wonderful

it's still the case even with now openzfs ? what do you trust now ?

it surpassed 30 junior engineers

image editing.. safety restrictions max level.. welcome to EU

or usecase/adoption increase ? also : capex!=operating costs

why don't let taiwan decide if they want to be ruled by ccp instead ?

usefulness is bound to scope/purpose, even if innovation stops, in 3y (thanks to hw and tuning progress ) when 4o costs 0.1$/M and 4.5 1$/M even being a small improvement ( which is not imo ), you will chose to use 4.5 , exactly like no one now want to use 3.5

given new USA ai diffusion rules will mistral be able to survive and attract new capitals ? , I mean, given that france is top tier country

based on meta declarations , amd is better at customized inference than nvda ,

meta uses amd mi300x for all inference and which bought 173.000 gpu ( 40% of their total gpu count ) according to that report https://www.cnbc.com/2023/12/06/meta-and-microsoft-to-buy-am...

maybe If there was a function to disable android running in the background from the tv ( which is the default), the menus and other primary functions would run better, maybe it's possible.. but do really I have to hack the tv ? how many will do that ?

and as I've said, I'm not even confident in performing a firmware reset at this point given that I would probably get only the version released with the tv without any successive update ( anyway not a solution for everyone )

even disconnected from the network, however I have to buy an external device to bypass some functions ( which is a non optimal workaround ),

however the bugs are not solved, however the TV remains very slow and I do not even have the courage to do a reset of the TV sw because I do not know if I could reinstall the sw to a recent version (or it would remain with the first factory beta because the updates are deprecated anyway), it is a 7 year old 65" Sony Bravia for example (led, 4k, i'm good with video/audio I don't want to change it but i'm almost forced to.. ) , the menus are very slow, some recording and playback functions of content from the line crash (since the latest updates) and I often have to restart it.... all for a non-replaceable integrated soc of $50

talking about E-Waste , why modern tv doesn't have external device for the soc ? why I have to bypass the slow tv using a firestick , but the tv anyway become obsolete,

but the tv anyway becomes obsolete, and lacking security updates also potentially a danger for my home network and for the internet in general? in addition to losing some essential functions that only the soc integrated in the tv system can provide? would it have been enough to move the computational part containing the cpu, the memory and the sw on an external hw so as to be able to swap it (proprietary or not, but at least upgradable!)

if she wanted to use this sum to aggravate her addictions, it would be understandable why they don't want to accommodate her, also in light of the fact that for Sam it would certainly be less of a bother than dealing with this whole thing, if they are going down the most difficult path for them (the family)

it is possible that they are doing it for her own good, otherwise they would have given her even $1m a year, making her sign that all the accusations were false if they had been as perverse as some commentators in this thread suggest

she already said that more than 1y ago : https://www.eetimes.com/rocm-is-amds-no-1-priority-exec-says...

Rdna3 is not cdna

You imply that he knows exactly when which imo is not and could even be next year for what we knows.. Who know every paper yet to be published??

the article contain a quote from satya nadella : https://x.com/ryanshrout/status/1792953227841015897

fair? h100 NVL are two h100 in a single package.. which probably costs 2xh100 or more,

if so ok it's fair to compare 1 mi300x with 1 h100 NVL but then price ( and tco ) should be added to the some metrics conclusion , also the NVL is a 2xpci5.0 quad slot , so not the same thing..

I am not sure about system compatibility and if and how you can stack 8 of those in one system ( like you can do with non NVL and mi300x.. ) so it's a bit a diffent ( and more niche ) beast

mi300x production is ramping , in latest earning report lisa su said 1H2024 is production capped , 2h2024 have increased production ( and still have some to sell ), thanks probably to cowos and hbm3/(e?) supply improved

large orders for those accelerators are placed months ahead

meanwhile mi300x on microsoft are fully booked...

https://techcommunity.microsoft.com/t5/azure-high-performanc...

"Scalable AI infrastructure running the capable OpenAI models These VMs, and the software that powers them, were purpose-built for our own Azure AI services production workloads. We have already optimized the most capable natural language model in the world, GPT-4 Turbo, for these VMs. ND MI300X v5 VMs offer leading cost performance for popular OpenAI and open-source models."