- Synthetic
- Posts
- AI helps to clear landmines; The road to AGI
AI helps to clear landmines; The road to AGI
Plus, could AI replace your CEO?
Steve Brown
May 30, 2024
In partnership with
This Week in AI
Startup Safe Pro AI is using off-the-shelf drones and machine learning to automate the detection of antipersonnel mines, antitank mines, cluster munitions, and other horrors of war. Today, the main method of clearing mines is for people to use metal detectors carefully. Very carefully. Safe Pro AI is training an algorithm to detect buried and scattered munitions using a combination of thermal, multi-spectral, and visual sensors. Each year, around 5,000 people are injured or killed by landmines. It’s estimated that around 100 million active landmines are spread over 60 countries worldwide.
A new study of over 12,000 people in six countries conducted by Oxford University and the Reuters Institute has revealed that most respondents have never used tools like Copilot or ChatGPT with 20-30 percent saying they’ve never even heard of them. The study collected by YouGov questioned people in Argentina, Denmark, France, Japan, the UK, and the USA. Those who do use AI use it to answer factual questions, get advice, generate ideas, summarize text, write emails, get recommendations, source news, and seek support. While many respondents don’t yet use generative AI themselves, they have high expectations for how it will change social media, search, science, news, healthcare, banking, the military, politics, government, law enforcement, and retail.
Source: Oxford University, Reuters Institute, YouGov report, May 2024
It’s been a tough couple of weeks for OpenAI CEO Sam Altman. Last week, he was getting letters from Scarlett Johansson’s lawyers; this week, a board member he ousted in November 2023 (after he was temporarily kicked out of the company) is sharing her side of the story and accuses Altman of ‘outright lying.’ Recently, OpenAI co-founder and chief scientist Ilya Sutskever resigned from the company, lamenting that safety had taken a ‘backseat to shiny products.’ Enter OpenAI’s new safety and security committee whose first task will be to evaluate existing processes and safeguards and make recommendations to the board within 90 days.
In related news, Microsoft, Amazon, OpenAI, Google, Meta, Samsung, Anthropic, xAI, Mistral, Cohere, and IBM have pledged to publish AI safety measures for their foundational models. This agreement, announced at the Seoul AI Safety Summit builds on the Bletchley Agreement signed last year.
Video: Unitree Robot
This week we are showcasing another humanoid robot from China. What’s interesting about this one is that it costs just $16,000. Check out the video to see what the G1 can do for the money (1m 43s). Be careful with your nuts.
AI Tech and Innovation
Will humans ever build artificial general intelligence (AGI), AI as intelligent and capable as humans on all tasks? If so, how, and once we get there, how will we know it’s AGI? The pursuit of AGI is a race, with leading contenders OpenAI, Google DeepMind, Anthropic, and Meta each charting unique paths. The latest models are impressive but still exhibit significant limitations, and several key challenges must be overcome before we can even approach AGI including the development of System 2 thinking, episodic memory, world models, and better transfer learning and generalization capabilities.
The computing industry is famous for its constant back-and-forth battle between vertically integrated players (think Apple, currently) and horizontally integrated (the PC industry or Android phone sector). But what will happen in the AI world? Will big players go vertical to gain market differentiation, increased efficiency, and control over their stack (and thus their destiny)? Or will they take advantage of open-source development and the cost-effectiveness of horizontal market economics? The answer is probably both.
The performance of large language models seems to scale well with their size—their complexity, the number of parameters they have, and the size of the dataset they’re trained on. Scaling has delivered improved results in computer vision, natural language processing, and chatbots like Gemini and ChatGPT. Researchers are now debating whether scaling models will yield breakthroughs in robotics, with some arguing that there is not enough relevant data to train on and that robotics data is very heterogeneous—the control data for one robot’s body is totally different from another.
AI Insights
“We are nowwhere near the point of diminishing marginal returns on how powerful we can make AI models as we increase the scale of compute.”
Speaking at Microsoft Build, Scott then compared the compute used to train GPT-3 to that of a shark, GPT-4 to that of an Orca, and GPT-5 to that of a whale 🐳
“Might we all work for AI in the future?” asks The New York Times. A recent survey of CEOs and senior executives by Harvard University, MIT, and edX revealed that 47 percent of respondents believe most or all of the activities involved in the chief executive role should be completely automated or replaced by AI. A former IBM consulting partner believes “the change delivered by AI in corporations will be as great or greater at the higher strategic levels of management as the lower ranks.” While a human would likely still need to be in the CEO slot for legal accountability reasons, a 2017 survey of 1,000 UK workers found 42 percent would be “comfortable” taking orders from a computer. Another way to think about the future of leadership—supported by AI, anyone could become a CEO.
Freethink interviews a filmmaker, an economist, and the exec of an AI video production toolmaker to get their perspectives on how powerful generative AI like OpenAI’s Sora and Google’s Veo will transform how stories are imagined, told, and experienced.
Autonomous AI agents are probably the next step in AI evolution on our way towards full AGI—intelligent machine workers that can offload tasks and work with us to achieve goals. But what should that interface be, and what should be the division of labor? The author argues that AI product managers should learn from IKEA and cake mix makers (early cake mixes were a flop until they required people to add an egg) that humans want to maintain a sense of control and participate in the process, and thus copilots are more likely to receive widespread uptake than full-blown autonomous agents.
At Synthetic, we often hear from AI practitioners seeking practical advice, best practices, and tips on AI implementation. The AI in Production conference brings together leaders and engineers from GitHub, LinkedIn, Tesla, and other companies to share hands-on experiences and cutting-edge perspectives with product teams and software developers looking to gain actionable insights into building AI products. Use discount code SBROWN to get 20% off the regular ticket price.
Toolkit for the Future
Understand AI’s Risk and Opportunities
With the rapid growth of AI, it's crucial for businesses to understand both the risks and opportunities it presents. Discover how you can utilize AI to drive success while minimizing potential pitfalls.
In the MIT Artificial Intelligence: Implications for Business Strategy online short course you’ll gain:
Practical knowledge and a foundational understanding of AI's current state
The ability to identify and leverage AI opportunities for organizational growth
A focus on the managerial rather than technical aspects of AI to prepare you for strategic decision making
Focus on the email that matters. Deep Clean automatically identifies old, unimportant emails so you can delete them. Snooze emails until a better time and easily track emails that need a response. Automatically filter unimportant emails into specific folders. It works across every device where you check your email. Get to inbox zero with award-winning SaneBox.
Boast automates access to billions in R&D tax credits and innovation incentives so companies can fuel their growth while preserving equity and avoiding red tape. There are more than $20 billion in R&D tax credits available to innovative companies across North America. Boast is the only R&D tax credit solution that combines technical and R&D tax expertise with AI to simplify claim preparation and maximize returns.
Need fresh talent that can harness the latest AI tools to increase the speed and volume of marketing content creation? Unlock value with fully-vetted, top marketing people with experience in your industry. The marketing-as-a-service model allows you to hire talent on demand, part -time or full-time, in 30-day blocks of time.
Try this easy-to-use, AI-powered alternative to DocuSign to write proposals, create quotes, get eSignatures on contracts, build online forms, collect payments, and much more. Great for sales, HR, legal, finance, and IT. Free 14-day trial.
Create and sell online courses from your own website with this top-rated platform. Easy to use with hundreds of beautiful templates and a powerful generative AI assistant that makes content generation a breeze. The AI will create lessons, exams, assessments, and more. Free to try.
Relieve pressure on your support team and serve customers better and faster with a support automation platform. AI answers 90% of queries and escalates the rest to your team. It automatically updates its knowledge database so it’ll never escalate that same issue again. The platform supports SMS, social, email, and phone communication.