• Synthetic
  • Posts
  • AI is coming for remote workers first, says HBR

AI is coming for remote workers first, says HBR

Plus, AI talks to animals (and aliens) and how to avoid hallucintations

In partnership with

Subscribe to Synthetic

This week’s most interesting and relevant AI news and analysis

This Week in AI

The bromance between Jensen Huang and Mark Zuckerberg, CEOs of AI titans Nvidia and Meta, continued this week at SIGGRAPH 2024 when they sat down for an hour-long fireside chat. How could Jensen not love Zuck when he’s purchased over 600,000 leading-edge Nvidia GPUs? 😜 During the conversation, Zuck announced AI Studio, a product that makes it easy for creators, small businesses, or individuals to create custom AIs. He argued that AI characters and representatives will become ubiquitous and that “just like every business has an email address, a website, and a social media account….every business is going to have an AI.” Nvidia’s CEO concurred, as does Synthetic. Our view: In the 1980s, Bill Gates envisioned a PC on every desk and in every home. We believe there will be an AI to represent every brand, every organization, and every person.

Startup Friend is building a new AI device designed to provide companionship and emotional support to its wearer. The $99 device hangs around the owner’s neck like a pendant and requires no subscription. It pairs with their phone for internet access and sends text messages to communicate. You can review their “reveal trailer” here. First shipments are promised for Q1’25. Will it be the first successful AI-first device?

A team at Imperial College London has developed a technique to subtly mark written works so writers and publishers can later detect if their works have been scraped by AI companies for use in their models. The code to generate and detect these “copyright traps” is available on GitHub.

Quick Hits

  • Alphabet Under Investigation - A UK government corporate watchdog is investigating Alphabet’s partnership with competitor Anthropic.

  • Apple Shuns Nvidia - Apple will use Google TPUs and cloud infrastructure to power many of its AI features, avoiding Nvidia.

  • E.T., Talk to Our AI - SETI scientists suggest sending an LLM into space as a representative to teach aliens our languages and answer questions about humanity. 👽

  • America’s Aging Grid Can’t Handle AI - CNBC reports on AI’s growing power and water consumption and the challenge presented by long-term underinvestment in our electricity grid.⚡️

Video: How AI Could Help Us Talk to Animals

Will the 2020s be the decade that we learn to talk to animals? 🐘 🐳🦍

Vox’s excellent explainer video uses plain language to explain how we might use a combination of supervised and self-supervised learning to translate languages across species. It's worth a watch. (9 mins)

AI Tech and Innovation

AI Hallucinating

Galileo has released its latest report on AI hallucinations, comparing the accuracy of 22 open and closed-source models from Anthropic, OpenAI, Google, Alibaba, Meta, and others over short, medium, and long contexts. The report considers the accuracy improvements gained by using Retrieval Augmented Generation (RAG) with a vector database and compares the cost of using each model. They recommend the best-performing model, best performance for cost, and best open-source model for RAG applications.

One of the shortcomings of foundational models such as GPT-4o, Genesis, and Claude is their ability to solve complex math problems. Google DeepMind recently presented AlphaProof and AlphaGeometry 2, breakthrough models for advanced reasoning and problem-solving in mathematics. These advances are seen as important steps on the way to building artificial general intelligence (AGI). Human competitors in the math olympiad have 9 hours to solve six math challenges. DeepMind’s models solved one problem within minutes and took up to three days to solve other challenges. It correctly solved 4 of the 6 problems, equivalent to the silver-medal category. This achievement represents a major leap forward on the road to AGI.

Google DeepMind has announced three new additions to its open-source Gemma 2 family:

  • Gemma 2 2B - a new 2 billion-parameter model optimized for safety, performance, and efficiency. Google DeepMind claims it’s by far the fastest and highest-performing model in its size category.

  • ShieldGemma - a suite of safety content classifier models to filter the inputs and outputs of models and keep users safe.

  • Gemma Scope - a tool to offer insight into Gemma’s inner workings, aiding transparency.

DeepMind is a leader in AI safety and a strong advocate for taking a responsible approach to AI research and development.

As if the rewards for being the first to develop artificial general intelligence (AGI) weren’t enough, the new ARC Prize is designed to spur fresh thinking and overcome the challenge of building AI systems that can efficiently acquire new skills outside their training data, a key capability of AGI. ARC-AGI is a benchmark for an abstraction and reasoning corpus (ARC) introduced by AI researcher François Chollet in 2019.

“LLMs are not by themselves intelligent - they’re more like a way to store and retrieve knowledge, which is a component of intelligence, but it is not all there is”

François Chollet, Google Senior Staff Engineer

AI Insights

Looking for motivation to go back to the office? This might be it. Harvard Business Review argues that with the rise of autonomous AI agents like Devin, physical presence will differentiate the automated from the robot-proof and that if your job can be done remotely, it is far more likely to be one of the first to fall to AI task automation. A recent Upwork survey indicates that 22% of U.S. workers will be fully remote by 2025.

Gartner predicts that 30% of generative AI pilot projects will be abandoned by the end of 2025. The reason? Challenges include poor data quality, rising costs, and unclear business value. But there are many successes, too. Gartner surveyed early adopters of gen AI who reported an average revenue increase of 15.8%, 15.2% cost savings, and 22.6% productivity improvement with their projects. Gartner distinguishes between ‘everyday AI’ that boosts productivity gains and ‘game-changing AI’ that boosts creativity to yield breakthroughs, such as a pharmaceutical company discovering a new drug molecule with AI. Gartner advises taking a balanced approach in the tradeoff between internal infrastructure and model development versus token-usage fees for foundational models and between everyday AI and more ambitious, higher-risk, higher-return AI projects.

The Associated Press analyzes public statements by Vice President Kamala Harris and Former President Donald Trump on their stance towards AI and what policies we might expect from a future administration under their leadership.

Ready to revolutionize your workday with AI?

Discover the key to unlocking unparalleled productivity with HubSpot’s free guide to using ChatGPT at work. You’ll find practical insights, useful integrations, and 100 prompt ideas to help you unleash the power of AI for a more efficient, impactful professional life.

Toolkit for the Future

Maintain healthy calendar habits, improve your productivity, optimize cross-team meetings, boost collaboration, and improve work-life balance. Reclaim schedules 1:1s and defends your calendar so you can focus. Syncs multiple calendars and integrates with Google Calendar, Zoom, Slack, and HubSpot.

Need to expand internationally, but not sure where to start? Oyster makes it easy to find, hire, and retain local talent. They handle the details (country-specific labor laws, international tax laws, compliance, and global payroll) so you can focus on finding the right talent and ramping their impact.

Build intelligent personal shopping experiences by adding the Manifest AI chatbot to your site. Help shoppers find what they need, faster. Double add-to-cart and conversion rates and get 25% higher AOV. Easy Shopify and help desk integration. Free 14-day trials.

Hire top talent with industry-specific expertise to build world-class engineering teams and solve complex business problems. AI-assisted sourcing, vetting, and hiring deliver the right talent at the right time, globally.

Take customer support to the next level with powerful AI features including chatbots, workflow automation, emotion insights, knowledge bases, data-drive insights, real-time reporting, and empathy coaching to help your support teams increase productivity and customer satisfaction.

Incorporate your company, access one-click growth tools, stay compliant, and manage everything your business needs — all online, from anywhere. Launch your U.S. business in minutes with no paperwork or legal headaches.