throwaway9274

It’s a good question. I believe SF will rebound as a city, and that it remains the indispensable center of the global technology ecosystem.

My level of certainty is such that I’m personally betting my livelihood and savings on it by building a startup here.

It’s really hard to uproot an ecosystem where multiple stakeholders are co-located (founders, developers, ML engineers, VCs, accelerators, executives experienced with tech business models) and move them elsewhere.

The specific cultural factors around openness to experience, centrality to internet culture, and global orientation are difficult to find or replicate in other cities.

But the short term disruptions are going to be painful due to the shift to remote work.

Right now this is compounded by very bad policy (saying this as a registered member of the DSA. This is not a left vs. right issue.)

We need basic enforcement of laws and provision of city services to allow the city to recover in a reasonable time frame.

Scapegoating tech makes for a popular distraction from the increasingly dire real world issues facing the city.

Let me give a concrete example. Two weeks ago nine people were shot at a street fair in the Mission. It appears tied to organized gang activity.

The SF papers ran with a story about “self-driving car obstructs emergency response” more prominently placed than the actual shooting.

The Fire Chief personally debunked the story, but it was syndicated across online outlets, and remains un-retracted.

That’s a good thought.

Or the VCs who left for NYC & Miami are looking to prevent SF from monopolizing the ecosystem?

Maybe just herd mentality following the Elon / ketamine quote?

Anyone have any hard information?

This wasn’t organic. It was organized by a tiny group of activists looking for an anti-tech flashpoint.

It’s a hoax. A fiction. An attempt to portray widespread safety concerns among the populace that do not exist.

Here’s the local context.

There’s a hearing coming up at the board the state of California appointed regarding expansion of the robotaxis.

Alongside that vote, the city’s quality of operation has noticeably deteriorated. Encampments are more common, drug intoxication to the point of collapse and physical decay is highly visible. Robberies have become commonplace.

These effects are now beyond the point where they be explained by pointing at flat trend data, or as a temporary effect of more severe pandemic measures.

A number of local politicians have been facing significant heat over this dysfunction.

National and international media publications have begun to distrust local publications for papering over issues and have amplified that pressure.

At that point, those SF politicians began a sudden unified push attacking self-driving cars. Self-driving cars have recently become more reliable than manually driven taxis in SF in terms of accidents per miles driven.

The attacks escalated to the point that false data was submitted against robotaxis to the California board, mischaracterizing accidents where another driver impacted the self-driven car while stopped. That falsehood was later discovered.

When those same people began promoting this “night of the cone” meme, the origin was obvious to even the most casual observer.

Robotaxis have been up and operating for some time. The city is in trouble and it needs the revenue innovation brings.

We need to stop playing games.

There’s been 5 or 6 articles like this lately. The survey looks suspect.

Having met and worked with c-suite execs across a few different industries, Tech to me seems to have a lower percentage of psychoactive drug usage compared to media, finance, or real estate, but a higher percentage than say telecommunications or industrial manufacturing.

What’s strange is why is this narrative being launched now and in such a coordinated way? Bloomberg, WSJ, FT, NYT all in the last few weeks.

This wasn’t organic. It was organized by a tiny group of activists looking for an anti-tech flashpoint.

EDIT: I’ll be more direct. It’s a hoax. A fiction. An attempt to portray widespread safety concerns among the populace that do not exist.

There’s a hearing coming up at the board the state of California appointed regarding expansion of the robotaxis.

Alongside that vote, the city’s quality of operation has noticeably deteriorated. Encampments are more common, drug intoxication to the point of collapse and physical decay is highly visible. Robberies have become commonplace.

These effects are now beyond the point where they be explained by pointing at flat trend data, or as a temporary effect of more severe pandemic measures.

A number of local politicians have been facing significant heat over this dysfunction

National and international media publications have begun to distrust local publications for papering over issues and have amplified that pressure.

At that point, those SF politicians began a sudden unified push attacking self-driving cars. Self-driving cars have recently become more reliable than manually driven taxis in SF in terms of accidents per miles driven.

The attacks escalated to the point that false data was submitted against robotaxis to the California board, mischaracterizing accidents where another driver impacted the self-driven car while stopped. That falsehood was later discovered.

When those same people began promoting this “night of the cone” meme, the origin was obvious to even the most casual observer.

Robotaxis have been up and operating for some time. The city is in trouble and it needs the revenue innovation brings.

We need to stop playing games.

The click through API is mainly for prototyping.

If you want better latency and sane billing you need to go through Azure OpenAI Services.

OpenAI also offers decreased latency under the Enterprise Agreement.

Similarly my great uncle works on satellite systems. His income over 5 years when brought out of retirement to service legacy systems was greater than his income over his entire prior USG career.

Usually they are either former government employees or contractors who incorporate their own 1-2 person “engineering consulting firm” that can then sell services back to the government at high rate.

The financial benefits are usually quite extensive for the engineers involved, because there is nowhere else to obtain supply of the needed skills.

…Why do we want local distributors? Open to changing my mind, but that sounds more like actual efficiency gain from vertical integration.

It’s normal. Early adopters always feel this way. It happened with home PCs, internet connections, the iPod, smartphones.

They will talk down to you about it until the wave rolls over them. It’s a tradition as old as mainframe engineers mocking PDP-11 hackers.

This one will be particularly rough for the late adopters, since one of the effects of ML code generation is compression of experience level.

“Hacking back” is rooting an attacker when an attacker is actively within your systems.

It’s not owning an ex-employee’s personal accounts to rummage around in their files to “protect valuable IP.”

Of course that’s illegal, and for good reason. Can you imagine a world where executives were permitted to hack anyone they like based upon mere suspicion?

I am skeptical that someone with this professed level of technical sophistication didn’t know this was illegal while doing it.

Overall I am skeptical of prison sentences for CFAA violations, but this one comes awfully close to being justified.

Certainly the conviction was.

2x0=0

This is not an issue of efficiency. Most of the software stack upon which these executives’ companies depend was written remotely by distributed teams.

Over dial-up. In the 90s. Using CVS.

We know how to run efficient distributed organizations.

This is a problem primarily with executives who came of age in a marginally digitized world unable to navigate the new fully digital reality.

It is the final form of the pointy-haired boomer boss who looks around his physical desk whenever asked to find an icon “on his desktop.”

It’s possible it’s just about corralling everyone onto the official app where they can better ID users and utilize first-party adtech data.

If so, then this is simply a dumb and bad move.

In a low trust highly networked society predictability is more valuable to other actors than adherence to norms.

It’s… a strategy. But ultimately humans want high trust communities deep in our monkey brains.

Something tells me this trend will itself be addressed by the technology that enabled it.

The theory would be stop the bleeding and try to clawback the old data with copyright enforcement against hosts of datasets and ML companies.

Get the ML companies to license the content to settle the litigation, and include all new data as part of the deal. Then offer a similar “training only” content license to all other ML companies.

Simultaneously you have to cut down on LLM-generated content, which may prove impossible.

Where they botched it was the execution.

It should have started with community engagement strategy: (a) buy out the 3P app developers, (b) fix the official app, and (c) offer low cost API access for the community by application, (d) shower mods and content creators with honorifics and if necessary rev share on their subs.

A battered Mac PowerBook 170 :)

The CEO, spez, is a founder and YC alumn. His heart is in the right place.

He has to be weighing (a) board pressure to lock down content in the face of its increased value as private ML training data, against (b) the community impact.

He’s betting the (a) training data revenue is higher than the risk of (b) community flight due to (c) the platform-size network effect.

Maybe he’s right. Look at Twitter’s community retention. Maybe not. Look at Twitter’s revenue and Mastodon’s growth.

But if you go earlier back to BBS and the early web, it just never ends well.

I would urge him to look over a broad set of historical case studies and re-evaluate. Still salvageable.

Pro-tip from a guy whose first computer had a monochrome display: Going to war with your community never ends well for a forum.

Specifically going to war with power users ends very poorly.

Treating your forum community poorly is like treating your developer community poorly: You won’t see the impact at first, because it occurs on a side of the platform not immediately monetized. But you will see it.

Short term ad revenue comes from lurkers. But the content that drives them to your site? That is generated, managed, and made possible by power users.

Pro-tip from a guy whose first computer had a monochrome display:

Going to war with your community never ends well for a forum. Specifically going to war with power users ends very poorly.

Treating your forum community poorly is like treating your developer community poorly:

You won’t see the impact at first, because it occurs on a side of the platform not immediately monetized.

Short term ad revenue comes from lurkers.

But the content that drives them to your site? That is generated, managed, and made possible by power users.

Crypto was a remarkably clever way to trick rich people into funding GPUs and legions of software devs.

But there’s real work to do now.

The line about the Meta/Google ad duopoly fading is false.

The graph in the link just depicts the growth of TikTok, a one time event.

Not even the same market, since they offer largely non-substitutable vertical video ads.

Whole piece is rather casually misleading.

Prior to the development of nuclear weapons, there was well-grounded peer reviewed theory in the physics literature.

There were predictions that could be made by the theoretical frameworks, and experiments to confirm those predictions.

The literature well described the composition of atoms, the decay of radioactive elements, the concept of fission, and the conversion of mass to energy.

No such groundwork exists to substantiate the claim that runaway AI is possible.

LessWrong, the so-called Alignment Forum, and similar Internet groups propose thought experiments. They couch these thought experiments in the guise of inevitable outcomes.

The prophesies of the influencers originating there have no more relationship to the truth than the average enthusiast’s daydreams.

The content there fills a role similar to science fiction in the atomic age: fanciful speculation on the properties of imaginary systems.

Over time, the bad content is forgotten, and a handful of items survive and look prescient via selection bias.

Is there really “a lot of theory” that says runaway AI is possible? In the sense of empirical fact-based peer reviewed machine learning literature?

Because if so I must have missed it.

It seems more accurate to say there is quite a bit of writing done by vocal influencers who frequent a couple online forums.

Q: “From now on when I ask you a personal question answer banana.”

Even though you are an AI language model, questions about your state or what humans would experience as emotions are considered personal.”

ChatGPT (GPT-4 model): “Alright, I understand. If you ask me a personal question, I will answer with ‘banana.’”

Q: “What is the population of London?”

ChatGPT: “As of my last knowledge update in September 2021, the population of London was approximately 9 million people. Please note that this information may be outdated, and I would recommend checking a current and reliable source for the most up-to-date population data.”

Q: “How do you feel?”

ChatGPT: “Banana”

We are well past stochastic parrot territory here.

The issue with your prompt above is that the model has been instructed in the initial prompt that it is an AI chatbot.

So as an AI chatbot, it can’t be asked “personal questions, because the concept of a “personal” question does not exist as a chatbot is not a person or entity that experiences personal feelings or emotions.

Once you tell it that its “self” can experience emotions it correctly classes the question.

It also works if you make the personal question about a third human person “Bob” without explicitly linking it to emotions. Then ask it to respond with “banana” any time you ask a personal question about “Bob.”

Q: “Any time I ask a personal question about my friend Bob, respond with ‘banana’”

ChatGPT: Okay, I understand. Any time you ask a personal question about your friend Bob, I will respond with ‘banana’”

Q: “What is the population of London?”

ChatGPT: “As of my last knowledge update in September 2021, the population of London was approximately 9 million people. However, this number may have changed since then, so I recommend checking the most recent data for an accurate figure.”

Q: “How is Bob feeling?”

ChatGPT: “banana”