Author HubMaximilian Schreiner

Max is the managing editor of THE DECODER, bringing his background in philosophy to explore questions of consciousness and whether machines truly think or just pretend to.
AI research

New Othello experiment supports the world model hypothesis for large language models

AI in practice

ChatGPT might be draining your brain, MIT warns - what ‘cognitive debt’ means for you

AI in practice

Anthropic releases Claude 4 with new safety measures targeting CBRN misuse

Ad
AI in practice
Short

Agility Robotics, maker of the Digit humanoid robot, and logistics service provider GXO Logistics have signed a multi-year agreement to commercially integrate Digit robots into GXO's logistics centers. The agreement, which will follow a pilot in late 2023, represents both the industry's first formal commercial deployment and the first robotics-as-a-service (RaaS) deployment of humanoid robots, according to the companies. Under the RaaS agreement, GXO will deploy Digit robots alongside the Agility Arc cloud automation platform. At a SPANX omnichannel distribution center in Atlanta, the Digit robots are assisting with repetitive tasks such as moving totes and placing them on conveyor belts. The companies plan to explore additional use cases and expand the use of Digit as needed.

AI in practice
Short

According to Bloomberg reporter Mark Gurman, Apple is working to bring Apple Intelligence to the Vision Pro headset. One challenge is to optimize the features for mixed reality. The AI features will not be released for the Vision Pro until next year - Apple Intelligence will launch on all other supported devices in the fall. By then, Gurman expects a deal with Google or Anthropic to support additional AI models. Longer term, he speculates, the company may be planning a monthly subscription service like "Apple Intelligence+" that offers additional features to monetize the technology. Apple already takes a cut of subscription revenue from any AI partner it brings on board. "The company will be less reliant on hardware tweaks to drive its business and will actually be making money from AI — something everyone in Silicon Valley is hoping to pull off," Gurman says.

Ad
Ad
AI and society
Short

Microsoft founder Bill Gates is not worried about the rising electricity consumption caused by AI applications. Speaking at an event in London, Gates said that AI will ultimately help reduce energy consumption and accelerate the transition to sustainable energy sources. According to Gates, data centers would only cause an additional 6 percent of electricity consumption in the worst case, but more likely just 2 to 2.5 percent. "The question is, will AI accelerate a more than 6 percent reduction? And the answer is: certainly," he said. Gates also believes that the increasing demand for electricity will lead to new investments in green energy, as technology companies are willing to pay more for green electricity "to say that they're using green energy." However, studies by Goldman Sachs and the Electric Power Research Institute predict that data centers could account for up to 9 percent of U.S. electricity consumption by 2030. Critics worry that the AI boom could increase the use of fossil fuels.

AI and society
Short

The EU Commission is looking into a possible antitrust probe into the partnership between Microsoft and OpenAI after dropping a merger review. EU Competition Commissioner Margrethe Vestager said on Friday: "The key question was whether Microsoft had acquired control on a lasting basis over OpenAI. After a thorough review we concluded that such was not the case. So we are closing this chapter, but the story is not over." The European Commission investigated whether Microsoft's multi-billion dollar investment in OpenAI falls under the EU merger control regulation. According to Vestager, the EU is now investigating whether "certain exclusivity clauses" in the contract between Microsoft and OpenAI "could have a negative effect on competitors." The US and UK are also investigating the Microsoft-OpenAI alliance. The Commission has also announced that it will investigate Google's agreement with Samsung to pre-install Gemini on the company's smartphones.

AI in practice
Short

LMSYS Org has added image recognition to the Chatbot Arena to compare vision language models (VLMs) from OpenAI, Anthropic, Google, and other AI vendors. In two weeks, more than 17,000 user preferences were collected in more than 60 languages. GPT-4o and Claude 3.5 Sonnet performed significantly better at image recognition than Gemini 1.5 Pro and GPT-4 Turbo. While Claude 3 Opus is better than Gemini 1.5 Flash for language models, both are similarly good for VLMs. The open-source model Llava-v1.6-34b is slightly better than Claude-3-Haiku. The data collected shows common applications such as image description, math problems, document comprehension, meme explanation, and story writing. Next, the team plans to add support for multiple images, as well as PDFs, video, and audio. The Large Model Systems Organization (LMSYS Org) is an open research organization founded by UC Berkeley students and faculty in collaboration with UCSD and CMU.

Image: LMSYS
Ad
Ad
Ad
Ad
AI in practice
Short

At I/O Connect in Berlin, Google unveiled its Gemma 2 model series, which will be available to researchers and developers through Vertex AI starting next month. The model will be available in two versions: with 9 and 27 billion parameters. The open-source models are said to be more efficient and secure than their predecessors, and outperform models twice the size in terms of overall performance. At Google I/O in May, the company showed the first benchmarks of the large version, which was also able to keep up with Meta's Llama 3 with 70 billion parameters. According to Google, the new models are aimed at developers who want to integrate AI into their apps or edge devices such as smartphones, IoT devices, and PCs.

Google News

Bank details

IBAN: DE87 1203 0000 1086 0070 75
Account holder: DEEP CONTENT GbR
Purpose: Support THE DECODER
AI in practice
Update

Kimi-K2 is the next open-weight AI milestone from China after Deepseek

AI research

New Energy-Based Transformer architecture aims to bring better "System 2 thinking" to AI models

AI in practice

Musk unveils Grok 4 as xAI’s new AI model that beats OpenAI and Google on major benchmarks

Google News