falaki

No AI has been touching my text :)

I agree. The author is stretching several assumptions to get to his conclusions. The sub-title suggests he has solved the "Bronze Age collapse" riddle but his statement can simply be considered one of many factors already pointed out in the literature -- not the main reason.

The primary conclusion of the article are based on two fundamental assumptions:

1. Iron weapons did not exist before the "Bronze Age collapse." That is not borne by the evidence. Iron weapons came from the same centers and communities that used to produce bronze weapons. If these weapon makers were restricted to supply to power centers during the Bronze Age, why were they not restricted during the Iron Age?

2. Wide distribution of iron weaponry prevented formation of great powers after the "collapse". If true why and how were Assyrians and Persians able to build such large empires long after iron weapons were wide spread? Weapon distribution cannot be the primary explanation for why great powers did not form after 1200BC.

I have read Cline's work. He is not the original source of the ideas on bronze to iron transition. This is a very well studied topic in archeological metallurgy.

This is part of an essay I am working on the nature of technological waves. I will publish on my personal blog.

Some high-level observations about the story of Iron are helpful. First, Iron smelting was discovered independently in different parts of the world, but everywhere it was an evolution of the bronze smelting process. Second, the process took several centuries to evolve. The first iron pieces were considered bad bronze and discarded. When tin became rare the producers gradually increased the quality of the bad bronze. In all likelihood for centuries they did not recognize iron as a new metal but rather an improved “bad bronze” . The final notable observation of the story is who these inventors were. They were not the learned people of the ancient world. Rather they were artisans and laborers working in difficult conditions.

Long before the so-called “iron age” started, humans knew of iron. There are iron artifacts from the 4th millennium BCE. Humans were fascinated by these small pieces of iron, which were found in meteorites and cherished and valued them. However, usable iron was rare and extremely sought after to the extent that we know it was more valuable than gold or silver. How did humans invent the first technique to produce metallic Iron from ore?

The story of Iron starts with the production of Bronze. Bronze is an alloy of 90% copper. Depending on what constitutes the other 10% we get different types of bronze. The most common type which kick started the bronze age around 3500 BCE in western Asia is tin bronze. Unlike copper ore, tin is hard to find but a small amount of it can significantly improve the quality of copper and make it usable and strong. Ancient smelters produced bronze by melting the natural copper ore mixed with tin ore in simple kilns. The resulting bronze was the primary metal used in tools and arts for more than 2500 years.

Iron is a common impurity in copper ores. We also know that iron ore can, and was, used as a flux in the process of smelting copper. In simple terms, a flux agent is a chemical that attracts oxygen to itself and protects the primary product, which is copper, from oxidizing in high temperatures. Therefore, across the ancient world, bronze smelters had been encountering molten iron when they raised the kiln temperature high enough. However, they discarded the resulting metal because its copper impurity made it too soft and inferior to bronze.

Around 1,200 BCE the collapse of bronze age civilizations in the near east due to a “perfect storm” of drought, famine, disease, immigration and war affected trade routes that brought tin ore to the western Asia. During this time of bronze shortage some smelters seem to have resorted to using the lower quality iron for tool making. They had observed that adding carbon to iron makes it much harder but prone to breaking. We have evidence of iron tools built with layers of soft and hard iron to achieve desired outcomes. Over time iron smelters experimented with different amounts of carbon impurity and achieved desired levels of hardness and strength.

The tell is calling models "the AI" or "AI." That shows you the author has a fictional understanding of statistical machine learning and neural networks. To them it is a sentient being called "AI."

A must read for OpenAI engineers.

How many security incident reports and postmortems has Dwarfish Petal written in his career? Does he understand the basics of system security? Can he tell the difference between secure and insecure configuration? Paint me skeptical.

this ^

I cannot wait to have Astra analyze the discovery documents for this lawsuit.

The incident report leaves a lot to be desired.

- It was done by two institutes with organizational ties to OpenAI: METR and Redwood Research

- METR and Redwood Research are institutional pillars of the "AI Safety" wing of the Effective Altruism movement. This clearly shows their prior biases towards "AI existential risk" rather than technical/engineering root-causing of the incident

- If you reed the report, it is not on-par with what you find from other companies.

- The access that was given to both was mediated and controlled by OpenAI. It is not clear, if they were able to get to the bottom of engineering flaws. It is not clear if they could see all the audit logs, etc.

Considering all the above, I consider the whole episode more of a PR stunt. I understand that is not a majority opinion at this point.

They couldn't pick a more sinister headline for such an awesome technological development.

Fun to read this in light of this other post: https://news.ycombinator.com/item?id=49279627

I recommend using https://omnigent.ai over Grok Build or any other harness.

Yes exactly.

Also they suggest every company should build their own benchmark and repeat these tests with new models instead of relying on the SWE bench.

1) Many models are now competitive at the top tier, including open source. 2) GLM 5.2 in particular was a major step forward in open source coding agent performance, 3) Harnesses make a huge difference in cost-performance. 4) Cheaper per-token does not imply cheaper per-task.

The paper has some insights that may apply to AI coding: 1. The first companies that benefitted from the Dynamo were electrical light manufacturers. They were more intimately aware of the potential benefits and since they were new, they built the factories to take advantage of Dynamos. Think who is benefiting most from AI coding these days.

2. Existing manufacturing operations didn't realize that their factory layouts and processes were unsuitable for Dynamos. They would replace the power source but see no benefit from it. It took many years for old pipelines to depreciate and be replaced with new ones that were able to benefit from distributed mechanical power generation.

Absolutely worth reading.

I worked at Siri (post acquisition) 13 years ago as one of the early data scientists. Let's just say I am not a bit surprised.

There are a lot of Iranian-Americans in Silicon valley, and the broader tech. These people have family and relatives in Iran and not being able to contact has been extremely hard on them. If you have an Iranian colleague, please understand that they may not be able to perform and work as their usual. Hopefully this collective nightmare will end.

Some personal observations as I am in touch with a few folks inside Iran through Starlink.

1. The jamming/disruption is local to large cities most notably the capital, Tehran.

2. Even in Tehran it is not complete and my friends are able to send and receive messages. Uploading videos is harder.

3. The regime is now raiding homes that they suspect have Starlink terminals. I don't know how they identify them but I do wonder if they are using technology to locate them.

Some personal observations as I am in touch with a few folks inside Iran through Starlink.

1. The jamming/disruption is local to large cities most notably the capital, Tehran.

2. Even in Tehran it is not complete and my friends are able to send and receive messages. Uploading videos is harder.

3. The regime is now raiding homes that they suspect have Starlink terminals. I don't know how they identify them but I do wonder if they are using technology to locate them.

I left the comment on starlink on that thread. I should note some personal observations as I am in touch with a few folks inside Iran through Starlink.

1. The jamming/disruption is local to large cities most notably the capital, Tehran

2. Even in Tehran it is not complete and my friends are able to send and receive messages. Uploading videos is harder.

3. The regime is now raiding homes that they suspect have Starlink terminals. I don't know how they identify them but I do wonder if they are using technology to locate them.

Except that in the common law system of the United States, a judge can throw out the regulation.

Yes, there are NGOs and organizations that sponsor these and pay for them. Here is an example: https://www.forbes.com/sites/cyrusfarivar/2024/12/18/inside-...

I hear after the Ukraine war, Starlink became very good at thwarting jamming. I am confident the Iranians are not as sophisticated as the Russians in than front.

Fortunately, the government cannot enforce complete blackout because thousands of startlink terminals are active inside the country. They have been complaining about it [1] to no avail. Using these terminals activists and journalists continue to upload videos of demonstrations to social media which has enabled analyses that show demonstrations are very wide spread [2] and continue to grow.

[1] https://www.itu.int/en/ITU-R/conferences/RRB/Pages/Starlink....

[2] https://www.bbc.com/news/articles/cre28d2j2zxo