| | | 🔗 Stories, Tutorials & Articles | | | | | | | Breaking the scaling limits of analog computing | | | | | MIT researchers have developed a way to effectively scale an analog optical neural network, which could enable a super-fast, energy-efficient neural network that can function with the same accuracy as a digital one.
An optical neural network is composed of many connected components that function like reprogrammable, tunable mirrors called Mach-Zehnder Interferometers (MZI). However, errors that can occur in each MZI quickly accumulate as light moves from one device to the next.
The team developed a new type of MZI by adding an additional beam splitter to the end of the device, which makes it much easier for an MZI to reach the setting it needs to send all light from out through its bottom port. |
| | | | | | | | | | Scaling the Mastodon | | | | | | The article describes the experience of the author and their co-admin in scaling their Mastodon instance, chaos.social, in response to increased usage following a tweet from Elon Musk about buying Twitter. The article details the various steps they took to scale their setup, including adding additional servers and migrating the database to a separate server. |
| | | | | | | | | | How ChatGPT will change cybersecurity | | | | | It is important to remember that ChatGPT is still just a language model and should be treated with caution as it can generate nonsensical responses and refer to non-existent studies.
In the field of cybersecurity, ChatGPT can be used in various ways such as malware creation, analysis, vulnerability search, and security consulting. For example, it can be used by novice cybercriminals to create new Trojans and by InfoSec analysts to quickly explain what a piece of code does.
Additionally, ChatGPT can be used to generate convincing texts that can be used in spear-phishing attacks, which can increase the number of successful phishing attempts.
To combat these issues, content analysis experts are actively developing tools that detect chatbot texts. However, it is recommended to be vigilant and aware of the potential dangers of ChatGPT, as well as to learn how to spot bot-generated texts by looking for stylistic quirks and incongruities. |
| | | | | | | | | | Functional Programming Is Great. But It Ain’t Magic | | | | | According to the author, Functional Programming (FP) has become increasingly popular but it doesn't solve all problems. FP can be hard to understand and it's not true that functional programs don't have any side effects. FP helps write more reliable code but only if done right.
It's possible to write functional programs that have concurrency errors. FP doesn't handle concurrent scenarios for you, it just gives you tools to make decisions about handling side effects. Teams shouldn't lean on technology, thinking it will handle everything, because it will only do what it's told to do. |
| | | | | | | | | | The only kind of test I write | | | | | This post talks about the debate of different types of testing (unit tests, integration tests, end-to-end tests) and how the author only follows one kind of test.
The principles the author follows are: writing tests after a clear public interface is defined, emulating user behavior, not testing internal functions or state, focusing on coverage of use-cases, and breaking the rules if needed.
The tests should not include remote network calls or mocking internal code paths. The test should pass consistently with no network or internet connection and no state carried across different tests. |
| | | | | |
|
|
