🚀 Device lets you talk silently; 4 new amazing robot skills; Laser in war now?; AI realizes it's being tested & more
Is China a climate saint or villain? Can LLM help create bioweapons?
Hi,
This is Thomas, co-founder and CEO of digital agency KRDS, we have offices in Singapore, HK, Shanghai, Dubai and India. I’m also a co-founder of The WeChat Agency and Yelda.ai (more about me at the end).
You're receiving Future Weekly, my personal selection of news about some of the most exciting (and sometimes scary) developments in technology 🤖 summarized as bullet points to help you save time and anticipate the future 🔮.
First, you'll find small bites about many different news, and then further down these summaries:
The Mindblowing Experience of a Chatbot That Answers Really Instantly (main take-aways, from Wired)
The situation is bad, see the chart. We’ll see what could be the best solution to buy time in our fight against climate change.
Small Bites
Impressive 1-min demo: this wearable device lets one silently ask questions to a computer by articulating words internally and receiving answers through skull vibrations
With AlterEgo, users can surf the internet, ask questions, and receive answers through skull vibrations, without speaking aloud.
Cognition AI just demoed "Devin": an AI that can perform work usually done by well-paid software engineers (Wired)
Chatbots like ChatGPT and Gemini can generate code, but Devin went further, planning how to solve a problem, writing the code, and then testing and implementing it.
Devin generated a step-by-step plan for a project, generated code needed to access the APIs and run benchmarking tests, and created a website summarizing the results.
OpenAI built an expert panel to see if there is any danger of people using LLMs to create biological weapons. Short answer: no. It was all in Google anyway and knowing the theory of what to do is not the hard part. (OpenAI blog)
Sam Altman told some Korean reporters that GPT-5 is expected to exceed expectations (source, translation)
On March 14, Altman mentioned, "I am not sure when GPT-5 will be released, but it will make significant progress as a model taking a leap forward in advanced reasoning capabilities. There are many questions about whether there are any limits to GPT, but I can confidently say 'no'." He expressed confidence that if sufficient computing resources are invested, building Artificial General Intelligence (AGI) that surpasses human capabilities is entirely feasible.
"Many startups are happy assuming that GPT-5 will only make slight progress rather than significant advancements (since it presents more business opportunities), but I think this is a big mistake. In this case, as often happens when technological upheavals occur, they will be 'steamrolled' by the next-generation model."
It is virtually the first time CEO Altman has given such a confident 'signal' about the performance of GPT-5. He made it clear that building AGI is his and OpenAI's goal
Hold your breath! Business Insider shared: on track to put out GPT-5 sometime mid-year in 2024, likely during summer
Anthropic, developer of Claude, was founded by ex-OpenAI key employees, has raised more than $7 billion in a year, in part from Google and Amazon
The technology that underpinned ChatGPT was developed largely by Dario Amodei (now CEO and co-founder of Anthropic) and others who had worked at OpenAI before leaving to create Anthropic.
🤨Dario Amodei said there was a 10 to 25% chance that A.I. technology could destroy humanity. (NYT)
Also: Anthropic’s Claude 3 causes stir by seeming to realize when it was being tested.
Claude: "This pizza topping 'fact' may have been inserted as a joke or to test if I was paying attention." (ArsTechnica)
Meta proposes a new AI approach (Fast Company)
Yann LeCun proposes a new architecture (‘V-JEPA’) which, very crudely, comes down to getting AI models to watch a huge amount of video and infer cause and effect and physics from seeing what happens - in effect, what infant mammals do.
V-JEPA isn’t a generative model.
Next step is to add audio to the video, just like a child watching a muted TV then turning the sound up. The child would not only see how objects move, but also hear people talking about them. A model pretrained this way might learn that after a car speeds off a cliff it not only rushes toward the ground but makes a big sound upon landing.
Yann LeCun, to the objection "If a foundation model was worth much, it's a sure bet it wouldn't be open."
So explain to us why the entire software infrastructure of the internet is open source?
Why is the entire software stack of the mobile communication network open source?
These things used to be closed and proprietary and provided by commercial vendors.
They all lost to open source platforms because of market forces.
If it's infrastructure, the world demands that it be open source.
AI foundation are becoming a common infrastructure.
Demis Hassabis, on the future of AI, once we have agents, that is AI that can act and do stuff, not just answer questions:
"I've always said (...) that it is a big step change. Once we get agent-like systems working, AI will feel very different to current systems, which are basically passive Q&A systems, because they’ll suddenly become active learners. Of course, they'll be more useful as well, because they'll be able to do tasks for you, actually accomplish them. But we will have to be a lot more careful." (Wired)
For what it’s worth, the consensus prediction at forecast aggregator Metaculus sees “the first general AI system be devised, tested, and publicly announced” as happening in December 2031. Just two years ago, the forecast was 2057.
C'est mignon... "Minouche Shafik, who is now the president of Columbia University, said: “In the past, jobs were about muscles. Now they’re about brains, but in the future, they’ll be about the heart.”" (NYT)
Thomas: let’s see, but AI will get better and better at simulating emotions almost perfectly. The last bastion will likely be face-to-face human interactions in in real life…
Klarna, a fintech company, launched its AI assistant, in its first month it was able to handle 66% of customer service chats. That’s 2.3M chats which would’ve needed 700 full-time agents. (source)
🚀 The incredible pace at SpaceX (source)
Last week, SpaceX's Starship launched successfully and reached orbital velocity. Forget Mars, that launch itself largely validates its capability as an expendable rocket (that we won't try to recover)
Watch take-off here, and here for Starship re-entering Earth's atmosphere, with views through the plasma.
It is estimated that an expendable Starship could bring down SpaceX internal cost to low earth orbit (LEO) at around $500/kg (compared to $2000/kg with Falcon 9 in expendable mode, already record low compared to competitors, Ariane 64, not even launched once yet, is looking at more than $5000/kg to LEO, €115 million for 21,500 kg)
SpaceX COO Gwynne Shotwell said that Starship could fly again in about six weeks.
And earlier in March, in less than 24 hours, SpaceX teams in Texas, Florida, and California supervised three Falcon 9 rocket launches and completed a full dress rehearsal ahead of the next flight of the company's giant Starship launch vehicle.
It included the additional complexity of operating a Dragon crew capsule en route to the ISS, plus the Starship countdown in Texas. While all this was going on, a handful of ground controllers also monitored the health of the Dragon spacecraft currently docked at the space station.
The main constraint to launch now? Launch pad availability" said Matthew Dominick, the NASA commander of the Crew-8 mission. "We’re at a cool spot in spaceflight right now. We’ve got rockets competing for launch pads, so you’re not waiting on payloads. You’re not waiting on rockets. You’re waiting on launch pads now."
The company aims to launch its Falcon rocket fleet more than 140 times this year, up from 96 Falcon 9 and Falcon Heavy flights in 2023. SpaceX wants to launch "at least" nine Starship test flights this year.
🤖 Robotics department
Tesla's Optimus: see its progress in one 9 months in that 5-sec video
May 2023 (left) vs February 2024 (right). Both are real time 1x speed.
Following the announcement of its partnership with OpenAI, tech startup Figure has released a new clip of its humanoid robot, dubbed Figure 01, chatting with an engineer as it puts away the dishes.
The robot can:
describe its visual experience
plan future actions
reflect on its memory
explain its reasoning verbally
"I see a red apple on a plate in the center of the table, a drying rack with cups and a plate, and you standing nearby with your hand on the table," the robot said in an uncanny voice, showing off OpenAI's "speech-to-speech reasoning" skills.
"Can I have something to eat?" the nearby human tester asked it.
"Sure thing," it responded, handing over the red apple with ease.
"So I gave you the apple because it’s the only, uh, edible item I could provide you with from the table," the robot said when asked why it handed him the apple, sliding in an unnervingly human-like "uh" mid-sentence.
Phoenix : Video: High-speed humanoid feels like a step change in robotics
You've seen a ton of videos of humanoid robots – but this one feels different. It's Sanctuary's Phoenix bot, with "the world's best robot hands," working totally autonomously at near-human speeds – much faster than Tesla's or Figure's robots.
Phoenix's hands are hydraulically actuated as opposed to the electric motors in others like Optimus and the Figure 01. "We moved over to hydraulics," she says, "and there's disadvantages there, they leak and they're very expensive to do R&D on."
"But it's the only technology that gives us a combination of three factors that are very important: obviously precision, but there's also speed and strength."
"If you see a robot doing something really dextrous," she continues, "like threading a needle, or doing a button or something, you have to ask yourself, could that hand also lift a 50-pound suitcase? And could it also move fast enough to, say, type on a keyboard? Usually people only show one of these three factors, but the ultimate hand needs to have all three. Currently, they hydraulic technology's the only one that can get you all three of those."
Watch an AI Robot Dog master an Agility Course It’s Never Seen Before: 1-min video
This is very different from the usual Boston Dynamics videos which are impressive but typically involve humans painstakingly programming every step or training on the same highly controlled environments over and over.
One of the most impressive aspects of the research is the fact the robot was trained in simulation.
Chinese car maker BYD launches a 420km range electric car for under $15,000. (source)
CATL, a Chinese company that is the world’s leading battery storage maker, prepares to halve the cost per kWh of its lithium iron phosphate (LFP) cells by mid-2024. (source)
Elon Musk’s German Tesla plant suffers close to $1 billion in damages after attack by the ‘dumbest eco-terrorists on Earth’ (source)
More than 2 billion metric tonnes of rare earth metals were just discovered in the US, dwarfing China’s 44 million tonne reserves. (WSJ)
Sinopec, China’s oil refining and distributing giant, announced peak gasoline consumption in that country last year, in 2023, due to the massive electrification of ground transportation
So, Is China a climate saint or villain?, asks The Economist
Chinese companies make 90% of the world’s solar cells (the building blocks of solar panels), 60% of its lithium-ion batteries and over half of its electric vehicles. These industries are known as the “new three” in China.
Last year clean-energy industries accounted for 40% of China’s GDP growth,
China has become the laboratory for the world on green tech. Last year it invested in the field 38% of the world’s total and more than double the amount that was invested in America, according to BloombergNEF, a research firm.
WhatsApp-based AI maths tutor improves students’ scores in Ghana (scientific paper)
Another very intricate, complicated task robots are learning to do autonomously: suturing
A video taken by researchers at the University of California, Berkeley, shows a two-armed robot completing six stitches (points de suture) in a row on a simple wound in imitation skin, passing the needle through the tissue and from one robotic arm to the other while maintaining tension on the thread.
Will lower cost of surgery ultimately, even replacing surgeons at some later point
Analysis of 5M freelancing jobs to see what jobs are being replaced by AI (source)
Very noticeable decrease in writing, customer service and translation jobs in Upwork since the release of ChatGPT.
Web design, graphics design, software development and video production jobs have been the most resilient and have not see any drop in demand (and even an increase)
The vast majority of companies are currently not focusing on training or fine-tuning their own LLMs, based on the lack of any increase in data annotation or machine learning jobs
The most popular use case, by far for AI right now is in developing chatbots
DragonFire: How lasers could revolutionize the way militaries counter enemy missiles and drones (CNN)
The Defense Ministry says the DragonFire can precisely hit a target as small as a coin “over long ranges,”
The Defense Ministry put the price of firing a 10-second laser burst at around $13. In contrast, the Standard Missile-2 used by the United States Navy for air defense costs more than $2 million per shot.
....but: Rain, fog and smoke scatter the light beams and diminish effectiveness; laser weapons release a lot of heat so require large cooling systems; mobile lasers, mounted on ships or aircraft, will need battery recharges; and the lasers must stay locked on moving targets for up to 10 seconds to burn holes in them
Finding this interesting? Share it on Whatsapp, just click here ❤️
If yes, feel free to take 3 seconds to forward that newsletter to one person, or share it on Whatsapp clicking here, I'd be immensely grateful.
If that email was forwarded to you, you can click here to subscribe and make sure to receive future editions in your mailbox (many CEOs and startup founders are subscribers)
More to chew!
The Mindblowing Experience of a Chatbot That Answers Instantly (Wired)
Groq makes chips optimized to speed up the large language models (it’s unrelated to Elon Musk’s chatbot Grok)
The experience of using a chatbot that doesn’t need even a few seconds to generate a response is shocking.
I typed in a straightforward request, as you do with LLMs these days: Write a musical about AI and dentistry. I had hardly stopped typing before my screen was filled with a detailed blueprint for the two-act Mysteries of the Mouth.
It was something a clever theater kid might have handed in at the end of a full-semester course
It’s no longer surprising to get stuff like that from a chatbot, and Groq uses modified versions of several open source LLMs, from places like Mistral or Meta.
The revelation was how quickly The Mysteries appeared, fully developed, on my screen. It took all of a second. (OpenAI’s ChapGPT, which proposed a musical called The AIgnificent Smile took around four seconds.)
That speedy turnaround left me disoriented. When there’s a pause between prompt and output, the feeling is that some artificial brain is cranking away at the answer, which comes as the result of gobs of computation—a process similar to human thought but faster.
But when the answers just … show up, you get a different feeling. Was the musical there all along? Did Groq have a day pass to all possible versions of the multiverse?
Speed opens all sorts of possibilities. Groq developers are working on apps where you can jam with AI-generated music in real time
Critics who have looked at Groq’s scheme, which requires more chips per user than the standard approach, say it will be too costly to implement once it’s out of the demo stage.
ChatGPT Plus costs $20 a month—how much more would anyone pay for faster responses? But Ross argues that the lower operating cost of Groq chips will more than cancel out the initial expenses.
How to buy time in our fight against climate change?
The chart shared at the start of the email is scary, right?
"The North Atlantic has been record-breakingly warm for almost a year now," Brian McNoldy, a senior researcher and hurricane formation expert" (NYT)
My take:
While we have to stop emitting greenhouse gases asap and will ultimately (and even capture and sequester some that is already there), it will take some time before technology gets us there, and in the meantime, it's not like we can force India and China and the rest of the developing countries to stop their development. It's hard enough already in developed countries (Gilets Jaunes, anyone?)
But we can't wait, as that chart illustrates, so we may have no other choice but to consider geoengineering, and in particular injection of SO2 in the stratosphere, which will reflect back to space some of the heat coming in from the sun.
The good news is that risks associated with it have been greatly exaggerated.
We know a lot about the impacts of SO2 on different systems of the Earth. There are so many papers. Nature has been testing this for us for billions of years through its volcanoes.
Today we’re emitting 50x more SO2 than we should need to pause global warming (I kid you not!), but not in the right part of the atmosphere.
Releasing it way higher up in the stratosphere at 20km of altitude will be way safer and effective than what's happening now randomly in the troposphere due to pollution.
Crazily, the decrease of SO2 emissions in shipping due to new regulation may be the cause for the recent acceleration in ocean warming, some scientists said (NYT)
We would release it with huge biodegradable balloons that can take it right to the stratosphere.
Its qualities:
It's effective: one gram of SO2 offsets one ton of CO2 for one year
It's cheap
Today, with very small volumes, it costs the leading startup in the field, Make Sunsets, $0.28 to offset one ton of CO2 (ie, per gram of SO2) for two years
Obviously that cost will go down as production volumes increase, and you can help them by ordering some of these offsets already
Comparatively, most serious carbon offsets (that can be proven to be permanent, along other criteria), cost currently much more than $100/ton of CO2
It's controllable, it's not irreversible: its effect stops when we stop injecting it
So it's not the long term solution we need, but it buys us time in the short term!
Why haven’t we done it at scale yet? Because the conversation was stuck in the academic world, where out of excessive precaution, some academics wanted to study this more.
Here is the best deep dive I've read on the topic, by
of Unchartered Territories, he looks thoroughly and convincingly at all the objections.
Previous newsletters:
Robots now fighting robots; What that AI experimenter did in 59 seconds; Crazy text-to-video AI
Impressive AI & robot demos; what's biocomputing?; Yann LeCun and Sam Altman on AGI
How a missing gene in our DNA affected the Battle of Trafalgar and gave birth to the Mafia
Superhuman: see how much work a professor did in 30 minutes with AI, + more gems about the future
That's it for this week :)
If you made it until here, well, thanks a lot for reading this newsletter! A very simple way to encourage me to continue doing this is to take a few seconds to:
transfer this to one curious friend, or share it on Whatsapp clicking here
click on the little star next to that email in your mailbox
click on the heart at the bottom of that email
Thank you so much in advance!
Here to subscribe to make sure you get the future editions if this one was forwarded to you.
More about me
I cofounded KRDS right after college back in 2008 in Paris, we now also have offices in Singapore, HK, Shanghai, Dubai and India, we're one of the largest independent digital agencies in Asia. More here.
Watch our latest game showreel: At KRDS, we take pride in designing and developing games from scratch for brands and organizations, big and small! Gamification has always been part of our DNA, since our early days creating viral apps on Facebook back in Paris as the very first Facebook marketing partner outside of the USA!
I also run The WeChat Agency for the Chinese market (the Government of Singapore Investment Corporation, GIC, is a client)
I’m the cofounder of Yelda.ai, which deploys voice AIs able to answer customers and prospects calling your company on the phone using natural language.
I also write op-eds and do podcasts at times. Here are my latest articles and podcasts, and here my last episode on the Abundance Makers podcast, interviewing one of the most promising clean tech CEOs in the US.
For the French speakers, I’ve written more than 50 articles on the future of technology over the past years, all can be found listed here.
Have a great weekend :)
Thomas