We strive to inform readers not only about developments in the bitcoin industry but also about adjacent technology spheres — cybersecurity and the world of artificial intelligence (AI).
ForkLog AI has gathered the most important AI news from the past week.
- Montana signed a law blocking TikTok within the state.
- OpenAI released an iOS app for ChatGPT.
- Anthropic trained its chatbot to read entire books.
- Stability AI launched a platform with generative AI models.
- SLAIT launched an AI tutor for American Sign Language.
- Most notable AI deals of the past week.
Montana becomes the first state to ban TikTok
Montana Governor Greg Gianforte signed a TikTok ban. The new rules take effect on January 1, 2024.
To protect Montanans’ personal and private data from the Chinese Communist Party, I have banned TikTok in Montana.
— Governor Greg Gianforte (@GovGianforte) May 17, 2023
The SB 419 law prohibits companies from operating “within the state’s territorial jurisdiction.” The document also obliges app stores to block TikTok downloads.
In case of non-compliance, marketplaces face a $10,000 daily fine. Users of the app will not be punished.
Opponents have filed suit against the state. Plaintiffs say the act infringes rights guaranteed by the Первой поправке к Конституции США.
We filed suit last night challenging Montana’s unconstitutional ban of TikTok, on behalf of 5 TikTok creators. Lead counsel is Ambika Kumar, who represented other creators in securing an injunction of President Trump’s 2020 ban.
https://t.co/p67xeT8w4A— Davis Wright Tremaine LLP (@DWTLaw) May 18, 2023
They also argue that the Montana government lacks authority to enact laws related to U.S. national security or foreign policy.
OpenAI released an iOS app for ChatGPT
OpenAI launched the ChatGPT app for iOS. The service will soon be available on Android as well.
The app is available free for users in the United States. It can sync history across devices and supports voice input via the Whisper speech-recognition system.
ChatGPT Plus subscribers will receive exclusive access to GPT-4 features, early access to capabilities, and faster response times in the app.
Media: Apple tests the M3 Pro chip
Apple has begun testing the M3 Pro chip on next-generation MacBook Pro computers running macOS 14.
According to Bloomberg’s Mark Gurman, the processor includes a 12-core CPU (six performance cores and six efficiency cores), an 18-core GPU, and 36 GB of RAM.
The company is expected to release the first MacBooks with the M3 Pro by the end of 2023 or early 2024.
Tesla publishes video featuring Optimus robots
Tesla has published a video showing its Optimus humanoid robots, demonstrating the latest developments from the robotics division.
Engineers have taught five units to walk slowly and to recognise and remember their surroundings. They have also trained them to detect and manipulate objects of various shapes by tracking human movement.
PaLM 2 reportedly contains 340 billion parameters
Google used five times as much data to train PaLM 2 as for its predecessor.
Journalists from CNBC studied internal company documents. They found that PaLM 2 was trained on 3.6 trillion tokens of text data, and that the model contains 340 billion parameters.
By comparison, the first version of PaLM was trained on 780 billion tokens and used 540 billion parameters.
Anthropic taught the chatbot to read books
Anthropic expanded the context window for its Claude chatbot from 9,000 tokens to 100,000, roughly equivalent to about 75,000 words.
According to the company, the algorithm can process and analyse hundreds of pages of documents or entire books, answering questions that require a synthesis of knowledge across multiple parts of the text. Users can also converse with the bot for hours or days before the system “forgets” the conversation history, the startup notes.
During testing, developers uploaded the book The Great Gatsby with one altered line. They wrote that Mr. Carraway was “an engineer working on machine-learning tools at Anthropic.” The model flagged the discrepancy with the original work in 22 seconds.
Amazon to integrate ChatGPT-like dialogue into its online store
Amazon plans to embed a ChatGPT-style conversational AI into its online store.
The tech giant has posted ML engineer vacancies. They describe “reimagining Amazon Search” via a new interactive dialog interface. The AI is expected to help answer product questions and to compare and recommend items.
OpenAI opens plugins in ChatGPT for subscribers
OpenAI canceled the waitlist for access to plugins in ChatGPT for Plus subscribers.
Enabling experimental features is possible via a new beta panel in settings. They allow ChatGPT to access current data from the internet and use more than 70 third-party plugins.
Stability AI launches a platform with generative AI models
The Stability AI launched the StableStudio platform. It is an open version of DreamStudio.
The service combines a text-to-image generator Stable Diffusion XL with a suite of large language models, StableLM. Later the company plans to add the chatbot StableVicuna, which uses reinforcement learning from human feedback.
Stability AI also announced a repository of plugins for StableStudio.
Intrinsic, a subsidiary of Alphabet, introduces a platform for programming robots
The Alphabet-owned company Intrinsic presented the Flowstate platform, enabling applications for industrial robots.
The software helps specialists with no prior exposure to develop workflows for hardware systems. CEO Wendy Tan White said the platform lets users design, build and deploy solutions.
At its core lies a graphical tree. It enables users to assemble more complex workflows. Existing skills include pose estimation, manipulation, insertion with applied force, and path planning.
Flowstate is in beta. Interested readers can apply for access at Intrinsic beta.
SLAIT launches AI tutor for American Sign Language
The SLAIT company launched an AI tutor for American Sign Language. It is accessible via a browser now, with a mobile app planned for later.
The SLAIT School service uses sign-language recognition technology and provides instant feedback on words or phrases. Learning is gamified.
A SLAIT School subscription costs $39 per month, but the first few lessons are free.
The most important AI deals of the past week
From May 14 to 20, AI startups attracted over $107 million. Here are the most interesting deals.
- The quadruped robot maker ANYbotics attracted $50 million.
- Together raised $20 million to develop open-source generative AI algorithms.
- Union AI raised $19.1 million to build a cloud platform for automating AI workflows and data management.
- Mitiga Solutions, a climate-tech startup raised $14.4 million.
- Amini attracted $2 million to support a platform for collecting and processing environmental data in Africa.
- Figorr, a cold-chain startup raised $1.5 million to scale IoT devices for monitoring storage conditions of perishable goods.
Also on ForkLog:
- Tesla cars taught to recognize driver drowsiness.
- China launched a campaign against AI-generated fake news.
- Sam Altman urged the U.S. to regulate AI.
- Elon Musk called himself the reason for OpenAI’s existence.
- CNET journalists founded a union to fight AI authors.
- 61% of Americans named AI an existential threat to humanity.
- American students had degrees revoked for using ChatGPT.
- The court upheld the algorithms of YouTube and Twitter in a terrorism case.
- Rishi Sunak will raise the topic of AI risks at the G7 summit.