Idk/idc about the "daily 0 days" that you referenced, I do care about being able to know with absolute certainty that my photos will be there 20 years from now.
599 karma · joined May 10, 2018
Idk/idc about the "daily 0 days" that you referenced, I do care about being able to know with absolute certainty that my photos will be there 20 years from now.
The national app store you get to use (the region of your iphone) is based on your billing address + card details (which obviously are validated when you enter them), not your phone location.
Page 9 of this recently published paper[1] is a strong indicator of how far non-US firms go to formally analyze and factor in these bandwidth constraints in building large models.
+24 hours - Nadella announcing the creation of an internal OpenAI and automatic employment offers to all OpenAI employees. Seems sensible. Microsoft's investment in OpenAI entitles them to 75% of all of OpenAI's net profit until they make their whole $13B investment back, not to mention retain ownership of 49% of OpenAI. Too much capital to play chess with, the division/auto-hire announcement still feels like it must have been done with a lot of hesitation and picking the least awful option, but what else? If the conclusion is that OpenAI's board is prepared to sabotage the company -- and the exceptionally vague (=> bullshit) reasoning OpenAI published for his termination supports that conclusion -- there's no other option.
But +48-72 hours, this is perhaps when the "holy shit" moment happened -- Bing's lifelong irrelevance as a consumer product and sudden ascension to relevance as Google's defense in an extremely serious lawsuit, and at least Google organically beat the other search engines.
The Microsoft board realized hemorrhaging OpenAI, whilst also probably owning xx% of all H100's (arguably the defining competitive advantage of GPT-as-a-Service co's), would almost definitely lead to them being in Google's predicament years down the line. By then, the FTC will not care that the "trigger" all those years ago was a board that became frantic about AGI, with Microsoft merely rescuing a ship that self-sabotaged.
So the chess started here, Ctrl+Z'ing everything, the overtly outward displays of compassion, some genuine, some perhaps encouraged, and everyone - including Sam - is back on the same page that Microsoft needs to back off, but crucially it has to appear as a reunification stemming from within OpenAI.
It is precisely at this point, in this chain of events, that the opportunity emerges for a OpenAI board replacement, with Sam being firmly on it, that is in line with Microsofts deep (arguably trauma-laden) desire to finally be the dominant player in something other than desktop operating systems.
No source, I doubt there will ever be, but this speculation of events, to me, lines up most with the actions of the parties involved.
5 occurrences of "Microsoft" in the article body vs 9 "OpenAI".
To make my point clear; this weirdness was definitely caused by the OpenAI board deviating from everybody's plans, including a member of the board themselves.
From there, Microsoft strategically reacted to a black swan event. From MS's pov, this was a zero-day attack that would manifest itself in 2-5 years, Google vs FTC style, and the mitigation strategy was providing the appearance of welcome arms++ to the tyranny of a board that makes kneejerk decisions, and having those open arms be rejected by the (natural) out-pour of sentiment.
The end result is identical, and Microsoft turned a zero-day attack into an irrefutable defense for any future FTC lawsuit.
No ignorance on it’s order-of-magnitude lethality multiplier over cars. But there is a certain cognitive-adrenaline fused pleasure that comes with seamlessly gliding your way through the perfect Bézier-esque curve over a gentle highway turn, whilst being held in the ever-present protective arms of gyroscopic precession, that makes the danger of it all worth it to me.
Rollercoasters let you experience a far greater degree of adrenaline and wind rushing through you, but don’t come close to this particular feeling of smiling at death.
1) do you acknowledge prompt engineering is a real skill set?
2) are you willing to improve your prompt engineering skill set through research and iteration?
There is much to learn about prompt engineering from that “Linux VM in ChatGPT” post and other impressive examples (where the goal of is to constrain ChatGPT to only engage in a specific task)
I replaced “Linux” in the prompt with “an Ubuntu instance in us-East-1 SSH’d into a Debian instance in us-East-2 ssh’d into a raspberry pi in germany……..” and at 4-5 levels deep it would forget the previous link when closing the “ssh connection” and go back to the first instance.
I also thought the sentence towards the end of the prompt about “when I want to tell you something in English I’ll do it like {this}” was redundant, but removing it leads to some weirdness like “cat file.txt” printing out the man page for cat 75% of the time.
md5sum / sha256sum also never work if the file contents are remotely non-trivial. so that seems to be the ceiling.. at least for this prompt :p
>>> spacex.current_stage 1
>>> spacex.current_time 1.10 min
>>> spacex.abort() Abort sequence initiated
>>> ls I'm sorry, but as a large language model...
>>> pwd I'm sorry, but as a large language model...
[0] https://www.dol.gov/agencies/eta/foreign-labor/wages#:~:text....
In the USA, for most work visas, you are given a 60 day death clock pleasantly sugarcoated as a 'grace period' to find a new employer or leave the country. Cue trying weave in that baggage in between screening calls & technical interviews to explain to prospective employers how you have 47 days and 15 hours left to be hired or leave. The complication of dealing with USCIS and attaining a work visa (usually bound to your employer) is a marathon even with unlimited time, but in a puny 8 week window?...
In my own experience this also happened in mid 2021 when Australia's borders were closed (to AU citizens as well...) so leaving back home wasn't even an option.
I don't know about you, but blasting my Spotify likes on a BT speaker in a crowded, and particularly an enclosed environment, would not be an enjoyable experience. I can imagine it's only enjoyable if one is sufficiently self absorbed enough to think that their music is universally enjoyed by everyone.
Interestingly, this leads to a startling scenario if you use an 'advanced' client such as Electrum, where you have the ability to create & sign a valid transaction, before broadcasting it. During that time, which is ofc defined entirely by your decision to hold off broadcasting the TXO, the transaction 'exists' but isn't recorded in the blockchain. In fact anyone who can get ahold of that transaction data (not to be confused with your private key) can send it off to the network, and it will be registered in the next block. So a valid transaction, once signed by the corresponding private key, is almost entirely removed from the signor, and the decision to broadcast it can be viewed as a seperate one altogether from signing.
Dumb TV is unambiguously a selling point now.
For the remaining 95%… I’ve configured it to delay doing anything until 10000ms passes so it doesn’t get in the way
https://nasa.github.io/fprime/Architecture/FPrimeArchitectur...
Care to share an example?
TIL tqdm now has a dead simple telegram hook https://tqdm.github.io/docs/contrib.telegram/