Tooploox CS and AI News #43

  • Scope:
  • Artificial Intelligence
Tooploox CS and AI News #43
Date: July 5, 2024 Author: Konrad Budek 6 min read

The moment the duck enters the AI race. 

This issue features the end of the AI Drive Thru, as tested by McDonalds, new Ad features from TikTok, and OpenAI’s first acquisition. 

06.03.2024 Amazon introduces Project PI to spot damaged products before shipping

Amazon’s Project PI uses generative AI and computer vision to inspect products for damage or incorrect color/size before shipping. Items are scanned in a tunnel and any defects are flagged and isolated for further evaluation. Active in several North American warehouses, Project PI aims to reduce returns and improve customer satisfaction. Human employees review flagged items to decide if they should be resold or donated. Additionally, Amazon is developing an AI tool to analyze customer feedback and investigate issues using multimodal large language models. This technology could help sellers identify instances of mislabeling and other problems.

More can be found on the Verge.

06.03.2024 McKinsey: Investments in GenAI start to bring value

A McKinsey survey reveals that investments in generative AI (GenAI) are starting to create value for organizations. Since the launch of OpenAI’s ChatGPT, 65% of organizations now use AI regularly, nearly double of last year’s figure. AI is increasingly adopted across multiple business functions, with notable use cases in marketing, sales, HR, supply chain, and inventory management. Organizations report significant cost reductions in HR and revenue increases in supply chain management. McKinsey classifies AI adopters into “takers,” “shapers,” and “makers,” predicting future hybrid ecosystems combining off-the-shelf, proprietary, and open-source AI models. Most organizations plan to invest more in AI over the next three years.

More can be found in this McKinsey report

06.05.2024 Now Asana can assign tasks on its own

Asana has introduced “AI teammates” to assist in task management by analyzing team history and skills to help assign work more effectively. The AI can identify and fill in missing information proactively. Built on Asana’s Work Graph, it maps relationships between tasks, information, and team members. Currently in testing, the AI helps marketing organizations create tailored content, translate assets, and standardize workflows. Asana also launched a chat interface for users to ask the AI questions about their projects.

More can be found on the Verge

06.06.2024 Google introduces Gemini to NotebookLM

Google’s note-taking app, NotebookLM, first launched last year, now allows users to upload Google Slides and web URLs as sources, in addition to Google Docs, PDFs, and text files. The new Notebook Guide feature can create study guides, FAQs, and briefing documents, with inline citations for fact-checking. Users can upload up to 50 sources per project, each up to 500,000 words long. NotebookLM, powered by Google’s Gemini 1.5 Pro, also enables users to ask questions about charts, images, and diagrams.

More can be found on the Verge.

06.07.2024 DuckDuckGo introduces

DuckDuckGo has launched, a platform for private AI chatbot interactions, featuring a clean user interface and offering four AI models: OpenAI’s GPT-3.5 Turbo, Anthropic’s Claude 3 Haiku, Meta’s Llama-3 70B, and Mistral AI’s Mixtral 8x7b. The service ensures privacy by not using user data to train models and stripping metadata to maintain anonymity. Saved chats are deleted within 30 days. DuckDuckGo aims to address growing data privacy concerns and may introduce a paid tier to reduce use limits. The service complements its search engine for a seamless user experience.

More can be found on AiNews.

06.12.2024 Luma Labs introduces Dream Machine

Luma AI, backed by Andreessen Horowitz, announced the free public beta of its AI video generation model, Dream Machine. The website was so popular that waiting times to test the system stretched long before the model could be put to use. 

The model can be tested on Luma Lab’s official website.

06.13.2024 Microsoft delays Recall due to security concerns

Microsoft is delaying the broad release of its Recall AI feature for Copilot+ PCs, initially set for June 18, in order to first preview it with Windows Insider Program members. The delay is to gather additional feedback and ensure the feature meets security and quality standards. Recall uses on-device AI to create a searchable visual timeline by capturing screen snapshots, aiding in productivity. However, privacy and security concerns have been raised about storing and analyzing such potentially sensitive data.

More can be found on VentureBeat.

06.13.2024 Google introduces Gemini for fitness

Researchers at Google introduced the Personal Health Large Language Model (PH-LLM), a version of Gemini fine-tuned for a time-series of personal health data drawn from wearables. This model provides superior answers and predictions compared to health and fitness experts. It uses generative AI to offer contextual, prescriptive outputs based on complex health behaviors. PH-LLM effectively analyzes data from wearables, overcoming previous challenges in interpreting and utilizing this information in health recommendations.

More can be found on VentureBeat.

06.14.2024 Meta halts plans for an AI assistant in the EU

Meta is pausing its AI assistant plans in Europe due to objections from Ireland’s privacy regulator, the DPC, which requested a delay in the training of language models on public content from Facebook and Instagram. Despite efforts to comply with privacy laws, Meta is disappointed and claims it can’t provide a high-quality product without local data. Meta continues to work with the DPC, noting that Google and OpenAI have used European data for AI training.

More can be found in the Verge.

06.17.2024 McDonalds ends AI Drive-thru trial

One of the world’s leading fast food brands will remove their chatbot-powered drive-thrus from over 100 testing locations in the US. The system, based on voice recognition and the generation of natural language responses, got its start in 2021 in cooperation with IBM. There was no official reason given for the trial’s end, yet media coverage has shown that the automated system frequently confused orders. 

More can be found in the Guardian

06.17.2024 TikTok introduces AI generated avatars

The new ad product introduced by TikTok lets brands use AI-generated avatars as actors in their ads. The system provides its users with a set of “stock” characters  depicting various ethnic and cultural backgrounds. The avatars can also speak multiple languages to deliver a better experience. 

Another ad product from TikTok enables brands to use AI-generated TikTok creators in their communication, for example, to have one speak the language of a target group. Content delivered in this way will be labeled as “AI generated.” 

More can be found in the Verge.

06.20.2024 Claude 3.5 Sonnet outperforms OpenAI and Google models

Anthropic has released its latest AI model, Claude 3.5 Sonnet, which outperforms OpenAI’s GPT-4o, Google’s Gemini 1.5 Pro, and older models across various benchmarks. This release marks a significant advancement in Anthropic’s innovation and enterprise-focused strategy. Claude 3.5 Sonnet is designed to revolutionize enterprise AI by offering enhanced performance and affordability. It demonstrates Anthropic’s commitment to rapid development and providing effective AI solutions tailored for business needs. The model’s superior capabilities make it a notable competitor in the AI market, emphasizing its potential for broad enterprise applications.

More can be found on VentureBeat.

06.21.2024 OpenAI acquires Rockset

The company’s first acquisition aims to use Rockset’s technology to boost the speed of AI products development. Rockset’s database enables companies to make better use of their data by boosting indexing and querying capabilities, to the point of enabling users to perform real-time analytics on their data. Apart from incorporating the tech into the OpenAI infrastructure, the Rockset team will join with OpenAI. 

More can be found in this official Blog post.  

06.25.2024 Google rolls out Gemini Sidebar for paying customers 

Google is beginning to roll out the Gemini side panel, enabling users to get quick access to Google AI. The tool can summarize incoming emails or suggest replies. The Mountain View company reminds its users that the AI can hallucinate and the content needs to be double-checked before sending. 

The company is also rolling out similar sidebars to use in Google Docs, Sheets and slides.

More can be found on the Verge

06.25.2024 Toys“R”Us unveils first commercial made with Sora

The iconic-yet-defunct brand of toy stores reminds users about its existence with an Ad that was made mostly using OpenAI’s Sora. The ad was created by the Ad Agency Native Foreign and, according to its creators, the content was ready much quicker than videos like this are usually delivered – within just a few weeks. According to Toys“R”Us, it is the first real ad to have used Sora commercially.  

More can be found on VentureBeat

06.28.2024 The Center for Investigative Journalism is suing OpenAI 

The company behind Mother Jones is suing OpenAI and Microsoft for an alleged copyright infringement. The company joins a growing group of news vendors and content creation companies that insist that OpenAI has no right to use their content to train Large Language Models.

More can be found on the Verge.

06.28.2024 YouTube looks for music deals with record labels

Youtube is looking into gaining permission to use music in their AI training processes and products. According to the Financial Times, the platform is willing to pay large sums of money to Universal Music Group, Sony Music Entertainment and Warner Records to legally train AI music tools using their content. The sums were not disclosed yet, but  sources claim that it will be a one-off payment rather than a royalty-based arrangement. 

More can be found in the Financial Times.