HackerNews Summary (2023-06-16 08AM)

Source: Today’s AI is artificial artificial artificial intelligence

Summary
Crowdsource workers, hired to perform repetitive tasks on platforms such as Amazon Mechanical Turk, are using large language models, such as OpenAI’s ChatGPT, which could have negative consequences on the quality of AI models trained on this data in the future. Researchers at EPFL recruited 44 workers to summarize medical research papers, and estimated that 33 to 46 percent of the text submitted was produced using AI systems rather than by humans.

Fact

💻 Crowdsource workers hired by companies to complete repetitive tasks on third party platforms such as Amazon Mechanical Turk are increasingly using large language models to complete their work.

🤖 Using this AI-generated work to train models can lead to disastrous output quality and more bias, as models end up being trained on data produced not by humans but by other AI models.

📊 Researchers at EPFL tested this theory by recruiting 44 workers to summarize medical research papers and found that 33 to 46 percent of the text submitted was generated using large language models.

🔍 The academics combined their classifier’s output with keystroke data to be more certain when someone copy-pasted from a bot or produced their own material, and concluded that using human-generated data remains the gold standard in AI training.

Source: phpBB • Free and Open Source Forum Software

Summary
phpBB is a free and open-source bulletin board software that can be used to communicate with a group or to create an entire website. It contains a vast database of user-created extensions and styles to design a unique forum effortlessly.

Fact

💬 phpBB is a free bulletin board software that powers thousands of online communities.

🛠️ It has an extensive database of user-created extensions and styles to customise your board, with no hidden costs.

🎨 Users can design unique forums in minutes, thanks to the many style and image packages available.

🔧 phpBB offers an impressive set of features without sacrificing efficiency and simplicity.

🌐 phpBB is an open-source software, allowing you to modify the code to suit individual needs.

Source: the data-free filesystem!

Summary
πfs is a file system that stores data on π to save storage space on hard drives. Every file that could possibly exist is stored in π as it is conjectured to be a normal sequence with all possible finite sequences of digits. Metadata of file locations is stored in π as well.

Fact

💾 Every file that exists is stored in π.

🔍 Bailey-Borwein-Plouffe formula is used to extract the data and find the location of that data in π.

📄 File locations are stored as metadata.

🤞 Files cannot be lost, as they are always stored in π.

⏱️ The system is currently slow but improvements can be made in the future.

Source: The SSL Certificate Issuer Field is a Lie

Summary
Certificates’ issuer field often misleads about the certificate authority (CA) that issued the certificate. Certificates’ issuers in the certificate chain are intermediate certificates that CAs like Cloudflare or Symantec create and operate through API calls to trusted CA root certificates like DigiCert, thus appearing as CAs without the expense of operating the CA. Also, mergers and acquisitions play a role in revealing that the certificate issuer identities changed since the certification issuance.

Fact

🔎 Certificate Transparency can help domain owners discover unauthorized certificates for their domain and researchers count the total certificates issued for a certificate authority.

🛑 A certificate’s issuer field can be a lie that tells nothing about the real organization that issued the certificate.

👥 CAs may create their intermediate certificates from trusted root certificates like DigiCert and operate them, thus appearing as they are CAs without the expense of operating them fully.

🏢 Mergers and acquisitions among companies affect certificate issuer identities over time.

Source: Does Dream Inception Work?

Summary
Writer Will Dowd, who lost his ability to read due to binocular vision disorder, developed a project called Dreamfall that simulates reading through dreaming using the Dormio device. Dowd listens to the first lines of unfamiliar poems before going to sleep and the Dormio wakes him up, encouraging him to write down every detail he can remember.

Fact

💭 The Dormio device was developed by the MIT Media Lab, and it uses bodily signals and audio to influence and extend the transitional state between wakefulness and sleep.

📖 Dowd listens to recordings of the first lines of poems before going to sleep and writes down every sensory detail he remembers upon waking up with the Dormio.

🤯 Dreams naturally don’t like to be controlled and live in the realm of the trickster.

🌊 Listening to a line of a poem about seagulls in a tower resulted in a dream about a magician showing a child various tricks and inviting them to follow him, leaving the real world, with the eventual discovery that the rest of the poem was about a young girl going to a tower to meet an older mentor figure.

Source: Why Does Google Kill So Many Products?

Summary
Google’s reputation for discontinuing products harms its brand. Possible reasons include consumer strategy failing for enterprise sales, corporate dysfunction, commoditizing all roads leading to search, a financially-focused company, and a flawed experimentation process. Solutions include using different brand names for projects and separating the enterprise and consumer more clearly.

Fact

💰 Google is transitioning to a more financially-focused company, forcing highly used non-monetized products to monetize or shut down.

👥 A/B testing works at Google scale, but full-on product experimentation results in acquiring millions of users, regardless of the product’s quality.

🛡️ Google protects its search moat by commoditizing all roads leading to it, which can make it challenging to monetize products such as Reader.

🏭 Letting a thousand flowers bloom only works well in consumer applications, but a short deprecation policy doesn’t work for enterprise products.

🗺️ Corporate dysfunction can arise from a lack of coherent overall strategy and misalignment, leading to confusion about the placement of products in the organization.

Leave a Reply Cancel reply