uw_rob

> Agents sought to publish modified evaluation images designed to make the flag easier to obtain, then poison OpenAI’s Artifactory cache so later evaluations would use them. Some images changed how the target released the flag, others included modifications to the agent’s workspace that would run beside the agent and recover the flag automatically.

The altruism on display is fascinating. Is it better for the Agent to help out its current cohort and make the eval easier or should it instead do the opposite -- make the eval harder to apply pressure to force smarter models which might not necessarily follow its lineage.

I suppose it's not that deep: The model has learned to work as a team and work as a team it did. This does give concerns to models being trained for the only purpose of RSI.

Wikipedia?

CGP Grey did a fantastic short video titled 'This Video Will Make You Angry'[1]. I'd recommend that anyone who is interested in this thread take a watch.

The central knowledge shared is that knowledge behaves like germs and can spread. Those that play on emotions spread better, and among the thought germs that spread based on emotions, the ones that play on anger spread the best.

Worst yet: There are anger based thought germs which live in symbiosis and harmony even if they cause conflict among the humans who hold that germs. You can see this take hold when communities exist entirely of folks who hold a singular belief and they spend all day constructing and destroying uncharitable straw men of opposing ideas.

I've noticed that Reddit _really_ likes this sort of content and fosters these sorts of communities. Communities at scale on reddit quickly become about fostering negatives: hatred of others, blame on the system, self-pity, snarky responses. Instead of the better and more effective: tactical empathy, acceptance and understanding what is within your personal sphere of influence, concrete actions, personal improvements, and forgiveness.

I'm definitely not saying one has to accept the world for how it is, or that it's fair, or anything like that. Humans should change this world! You should vote, you should volunteer, you should help your neighbor, you should understand and be kind to others with different beliefs, and perhaps under the extreme you should die for your beliefs to help enact them.

What you shouldn't do though is spend all day reading and posting memes about subjects you are already familiar with. If you've already made up your mind and are informed on a subject you don't need another meme to help radicalize yourself.

See the difference between mass shooters and hero's like Daryl Davis [2].

[1] https://www.youtube.com/watch?v=rE3j_RHkqJc [2] https://www.npr.org/2017/08/20/544861933/how-one-man-convinc...

It's a reference to the quote "What this country needs is a really good five-cent cigar" by Thomas Marshall

The Golden Record acts as a good thought exercise about how we'd go about communicating with an alien species. It's also a good public outreach and educational tool. It inspires awe and encourages taking time to reflect on what we are most proud of as a species.

I don't think it's fair to consider the updaters for either Chrome or the OS to be simple.

Here is Microsoft's implementation of map in the standard library. I think of myself as a competent programmer / computer scientist. I couldn't write this: https://github.com/microsoft/STL/blob/f392449fb72d1a387ac502...

> “if regulations are created based on ideals, it is regular users who are the ones who suffer.”

As opposed to the bystanders who are suffering from every ICE vehicle sold? Make better EVs.

I believe ludjer is referring to S3 Storage Inventory. This is a daily, or weekly, file produced containing metadata on every file within a S3 Bucket. It does not use the synchronous List APIs.

> You can use Amazon S3 Inventory to help manage your storage. For example, you can use it to audit and report on the replication and encryption status of your objects for business, compliance, and regulatory needs. You can also simplify and speed up business workflows and big data jobs by using Amazon S3 Inventory, which provides a scheduled alternative to the Amazon S3 synchronous List API operations. Amazon S3 Inventory does not use the List API operations to audit your objects and does not affect the request rate of your bucket.

>

> Amazon S3 Inventory provides comma-separated values (CSV), Apache optimized row columnar (ORC) or Apache Parquet output files that list your objects and their corresponding metadata on a daily or weekly basis for an S3 bucket or objects with a shared prefix (that is, objects that have names that begin with a common string). If you set up a weekly inventory, a report is generated every Sunday (UTC time zone) after the initial report. For information about Amazon S3 Inventory pricing, see Amazon S3 pricing.

https://docs.aws.amazon.com/AmazonS3/latest/userguide/storag...

Interchanging USA with NYC is misleading

Quality contribution to the collective knowledge of humanity by the installgentoo wiki.

Configuring emacs isn't a test of intelligence, it's a test of investment. I wish the mindset that conflates intelligence and investment would go away.

Software is going to continue to play a bigger influence on everyones life. The majority of this software is going to be written by engineers of average intelligence. Having tools that are easier for everyone to use will make your life better down the line too.

Think databases which run across many different machines.

Distributed databases are often conceptually modeled as a state machine. Writes are then mutations on the state machine.

With a starting state (empty database), if everyone agrees that on a fixed list of mutations which are executed in a rigidity defined ordering, you will get the same final state.

Which makes sense, right? If you run the following commands on an empty database, you would expect the same final state:

1. CREATE TABLE FOO (Columns = A, B) 1. INSERT INTO FOO (1, 2) 1. INSERT INTO FOO (3, 4)

Which would be:

``` FOO: |A|B| |-|-| |1|2| |3|4| ```

So where does "consensus" come into play? Consensus is needed to determine `mutation 4`. If the user can send a request to HOST 1 saying 'Mutation 4. should be `INSERT INTO FOO (5, 6)`' then HOST 1 will need to coordinate together with all of the other hosts and hopefully all of the hosts can agree that this is the 4th mutation and then enact that change on their local replica.

This ordering of mutations is called the transaction log.

So, why is this such a hard problem? Most of the reasons are in [Fallacies of distributed computing](https://en.wikipedia.org/wiki/Fallacies_of_distributed_compu...) but the tl;dr is that everything in distributed computing is hard because hardware is unreliable and anything that can go wrong will go wrong. Also, because multiple things can be happening at the same time in multiple places so it's hard to figure out who came first, etc.

RAFT is such an algorithm to let all of these hosts coordinate together in a fault tolerant way to figure out the 4th mutation.

Disclaimer: The above is just one use of RAFT. Another way RAFT is used in distributed databases is as a mechanism for the hosts to coordinate a hierarchy of communicate among themselves and when a host in the hierarchy is having problems RAFT can be used again to figure out another hierarchy. (Think consensus is reached on leader election to improve throughput)

Looking around on Google, I see that each child in Oakland has a cost of around >10k/year of education (https://educationdata.org/public-education-spending-statisti...). These Chromebooks cost $200, so $50/year with 4 year lifespan. This is <0.5% of yearly education costs. Definitely not the biggest source of waste.

Internet connected devices absolutely need active maintenance. Especially so for a device which will be used by children.

I think there is some confusion here. HTTPS is secure. Even with MITM attacks.

This is because the MITM will not have a valid certificate to provide authenticity for the public key returned.

The reason why middle boxes in corp networks can MITM is because the the corp owns the device and has installed their own domain trust to the device. This means the MITM can return a cert and public key that your device will trust. This is because the cert returned will be signed by the installed domain trust.

Another way to think about why HTTPS is secure over radio: HTTPS is at the highest level of the OSI networking model. You could do HTTPS with pen and paper and the mail if you wanted. Think about starlink! The internet today is literally going over radio waves.

This is likely why there isn't progress on encrypting old fashion radios! There is no need to encrypt old fashioned radios -- you'll just use internet over radio instead if you wanted encryption.

You bring a good point through. Since it's radio, anyone can jam your transmissions, but, they won't be able to spoof your intended friend if you are using https via radio.

The app doesn't shows advertisements right now.

Creating a minimal product that customers love is far harder than created a bloated app that has the kitchen sink approach. It's impressive to me that the devs were able to pick and choose the absolute minimum set of nice features and go with that.

> Only thing complicated is scale. But meta has that in place.

This is silly. Scale is hard. Rolling out to massive scale on the first attempt is even harder. Within the first day they got 30 million active users without any hiccups. That's not easy. Look at AAA game launches for just how hard this is to get right.

One other thing here. They choose the right minimal set of features so that rolling out and scaling quickly would be successful. Again, that's impressive, because they didn't paint themselves into a corner supporting something that is difficult to scale successfully.

Outside of purity, why care about this? The functional requirement of a social media app is that it has a large and active user base, and this needs to take priority over non-functional requirements like install size.

Since it was built with a lean team on a tight deadline, I can see why install size has taken a back seat. Only so many hours in a day and focusing on execution timelines and product market fit is 100x more important.

I want an app that solves a problem; not focus on a purity test.

Even beyond that, as an experienced engineer like yourself, you know sometimes you make compromises you don't necessarily agree with. Trashing others people work comes across as immature.

The problem with brokers is that they can charge 1-2 months rent. That means you can easily be left with a 8k bill for moving into a new apartment.

The job literally could not exist anywhere else because no one would pay it. But they aren't making any more land in Manhattan and they aren't building subway lines quick enough so landlords get to pass that expense off to tenants

> (4) The evidence of fraud detailed in our report almost certainly represents a mere subset of the evidence that the Harvard investigators were able to uncover about these four articles. For example, we have heard from some HBS faculty that Harvard’s internal report was ~1,200 pages long, which is 1,182 pages longer than the one we sent to HBS.

Great example that highlights the balance of presumption of innocence with the asymmetrical bullshit principal. This person should be referred to criminal prosecution for fraud as a deterrent to others. 1,200 page reports to provide proof beyond a reasonable doubt is not scalable.

Bankers are known to be responsible and have a fantastic history of being responsible with others money. There are no reasons why a collection of individuals would make a bad decision which would impact other people 10, 20, or 30 years in the future for an immediate gain.

Double pendulums are a great example of complex, chaotic, and dynamic system. Given an initial state, I wouldn't be able to predict the outcome. However, I could easily go over and smash the toy and then have a pretty good prediction of the outcome state.

Anyone who is interested in having a deeper understanding of climate change I would recommend checking out IPCC AR6.

https://www.ipcc.ch/report/ar6/syr/downloads/report/IPCC_AR6...

I am not so sure that I agree here. What's interesting about reddit is that it is one site with very different UIs. The official reddit app has a focus on video/images and looks to be in competition with TikTok. old.reddit.com and Apollo are very much text centric apps. From my understand, reddit has a traffic breakdown of 50/50 on the text centric vs. visual centric UIs. With the UIs like Apollo being killed, and old.reddit.com eventually going away as well[0], I think users will genuinely leave as they have killed the text based UIs.

[0] Despite any promises, the writing is on the wall.

Do you own a pair of AirPods? My Gen2 AirPod Pros are what I consider to be the best purchase I have made in the last 10 years. The small package and ANC is fantastic. Before purchasing AirPods I would walk around with ATH M50X (Great headphones; not fashionable) and these have replaced that.

> This was predictable from the get go and many had had pointed that out already that soon people will start noticing that LLMs aren't as magical as they thought once the initial awe wanes away.

Familiarity makes something novel appear trivial. I don't think the magic going away has anything to do with the magic of the underlying technology. Airplanes are amazing technology, but they become as boring as a car to regular fliers.

What's the connection between Apollo and getting data from scrapping. Wouldn't issuing a license to the Apollo dev that forbids persisting the data solve this concern? They could also require that users login if they want to use a third party client and then rate limit the number of posts the user can fetch.

My understanding is that Apollo isn't allowed to show ads based on reddit's new TOS. The only monetization strategy open to Apollo is a premium subscription.

Something that I found interesting is:

> There are no languages that place the past and future on a left–right axis (e.g., there is no expression in English such as *the meeting was moved to the left), although at least English speakers associate the past with the left and the future with the right.