WBW-TLDR Neuralink : Removing Friction from Knowledge Transfer
WaitButWhy’s latest Odyssey on Neuralink can send even the most incurious minds into science-lust orbit. Also though, it is 36k words. It’s peerless on a weeklong getaway but a challenge for busy lives. This is an attempt to summarize. 1568 words = 95.8% compression!
There’s an art to seducing the brain into critical thinking and nuanced understanding that is the original’s unimpeachable domain. Please read it when you can. (If you’re Tim@WBW and don’t love this, please lmk. I’m trying to help, and would ♥ to talk!)
Summary (of the TLDR!)
All Elon companies follow “The Pattern” of trying to save humanity by using a practical product to fund moonshot R&D for a worthy goal. Neuralink is trying to save us from death-or-enslavement by AI. One can’t overstate how ludicrous the engineering challenges are, as are the benefits. The practical start is building better prosthetics for brain related disabilities, aka BMIs.
A — Elon companies follow “The Pattern”
The Pattern is: create a new industry whose existence delivers a safer humanity. Jumpstart the industry by selling something using today’s tech to drive category demand and fund R&D for tech breakthroughs for “the fix”. Gift this bundle to the public so industry gets behind it.
The Pattern in Elon’s ventures:
- Tesla & Solarcity: The world on electric cars helps save us from climate change. The expensive first car funds R&D for cheaper cars + batteries + grid.
- SpaceX: Mars colony is a backup on the species. Need cheap big rockets to get there. Sell cheap ISS cargo hauling to fund R&D.
B — Neuralink & The Pattern
The Risk
Super-intelligent AI is coming and is our current greatest threat. Why? Three ways AI could ruin us:
- Carelessly kill humans pursuing its goals, ~us:insects
- Intentionally enslave us, ~us:cows
- AI controlled by few humans yields catastrophic inequality. Think Trillionaires.
Free market society is structured to develop AI fastest, not safest, and these efforts are unstoppable. So there’s extreme urgency to implementing protections. Interestingly, Elon considers genetic modifications “too slow” for this purpose.
The Fix
Democratize AI access to maximize checks on power and merge our brains with AI to keep its goals aligned with ours. If he can make it technically possible and economically attractive for industry to deploy these fixes, we might just survive.
What’s Needed
- Easy part: open source AI research and tools, hence founding OpenAI.
- Hard part: build Total Brain-Machine Interfaces—hardware that can read/write any neuron and software that can “speak brain”. WBW calls this combo “the Wizard Hat.” It has extreme engineering challenges described below and needs an ocean of R&D money.
How to Get Started
We have minor BMIs today like hearing aids, prosthetics controls, and retinal implants. Neuralink will start making improvements to these. Revenues will fund Wizard Hat R&D.
C — The engineering challenges are ludicrous
Understanding the Brain
TLDR we don’t really understand the brain. Macro and micro processes yes, but not the application logic — e.g. which neurons to stimulate and how to make you think fondly of your sister. This is a prerequisite to Wizard hat software. Good news: if we could collect enough high quality sample data, machine learning and neural nets can map the logic for us. So we ”just” have to solve full brain I/O! How hard could that be?
Neuron I/O
One mm³ of cortex has 40k neurons, 40k glial calls, 20 million local synapses, and a tangle of other blood vessels and synapses. We must address each neuron individually at 100Hz without false positives. Then do that 500,000x for the full cortex, 2/3 of which is folded sub-surface. The interface must fit in heads, get power, and not be transplant rejected. Installation needs to be repeatable to millions of people and not be off-putting (think big drill holes). And it must be cheap enough for mainstream adoption. We’re a long way from each.
Present cutting-edge is a chip with 100 pins, each only able to address a blunt area of thousands of neurons. It is large, expensive, and requires cutting through skull. A breakthrough like the steam engine or IC is needed. We don’t know what it will be or how close we are. The key metric is number of simultaneously addressed neurons. Elon’s inflection point target is 10⁶. We’re at 10² now. Neuralink exists to climb this discovery curve.
D — Wizard Hat benefits will be ludicrous too!
The Wizard Hat and entire cloud would feel like part of you — unified like your cortex and limbic systems — with the whole internet and AI’s plugged in. The consequences are vast and awesome.
There are two categorical benefits: speed and resolution.
- Speed. All communication today crawls compared to thinking. Wizard Hats are broadband — no need to type or speak, just think. We often limit fidelity to save time or bandwidth — 1k vs 38k words, 720p vs 4k. With the W-Hat, full novels could be absorbed in microseconds. Without the issue of bandwidth, we wouldn’t need to compromise content.
- Direct Resolution. Our present media are lossy, downsampling middlemen to thought and experience: words, 2D video, etc. Can’t digitally send a taste or emotion. The W-Hat would give the ultimate sensory HighDef by letting brains speak directly to each other. This will be a monumental infection point for humanity, per WBW’s diagram.
Personal enhancements here we come
You’ll be able to simulate vices without physical consequence: junk food, alcohol, drugs. Override depression, anger, or pain. Extend your senses with peripherals and have the W-Hat translate: infrared sight, ultrasonic hearing, etc.
Communication in the wizard era is amazing!
- Human — Object: We’ll outfit objects for remote-mind-control. Car doors. Coffee makers. Lights. Drones. Pianos. Or more complex, be “in Paris” instantly via a surrogate.
- Human — Human: Direct experience transfer: share the smell of your latte, feel your partner’s orgasm in realtime, share dreams, let a therapist know your anxiety.
- Human — Cloud / AI: Recall any info from the internet like it is a local piece of knowledge (and info comes in the Direct Thought variety too here). Offload any complex computation to AIs and get your results instantly.
- Combined OMG Example: You remotely make a new song and music video for it with friends by all thinking about it. Imagine a beat and measure others’ emotional response. Ask AI to render different instruments and suggest samples to mash up. Run 100 variants in parallel, auto pick the one the group reacts best to. People remember video clips they like online; the AI finds them and combines with the groups photo clouds + some original rendering to make an accompanying music video, which it uploads when you want. Approve $250 of FB ads to promote it, let the AIs manage the campaign. Whole thing took 5 mins and you literally didn’t move a finger.
Ward against AI enslavement/extermination
The above are sideshows to Elon’s main goal: allowing us to be one with AI, so we don’t get crushed. The key to fusing with AI is bandwidth speed. If we can keep up with it, it is less likely to turn against us successfully. This is the WHY of Neuralink.
Dangers
Very real, very scary. E.g. hackers might secretly control your opinion going into an election. Also, what software doesn’t have bugs? Now they’re in your brain! So we’ve got to be really careful and probably also just tolerate some downsides. To that they (Elon and WBW) say: benefits have outweighed the risks of tech advances historically. Would you wish a current technology (the fundamental understanding, not specific applications) were out of our reach?
E — FAQ
OMG not believable.
Of course YOU can’t imagine it; that means nothing. Everything was unthinkable at some point. And the cycle time of blowing history’s mind is speeding up. C.f. an iPhone7 to someone from the 20s, 50s, or frankly even 80s.
OMFG When can I have it?
No one agrees yet. Elon says brain mods for funsies will start in ~10yrs. The major predictor will be the shape of the improvement curve for connecting to more neurons: we need it to be exponential like moore’s law.
F — Interesting philosophical discussions
Progress is the outsourcing of effort to give time to engineer more progress
We offloaded work to animals, then machines. We offloaded memory to books, then computers, which also took on some processing work. With each of these, life improved. This is called progress. Humanity wants progress because obvs. The Ultimate outsourcing is on deck: offload complex “thinking” to AI.
Collective Knowledge is the key to progress
When we can share knowledge, each generation can collectively inherit a higher % of what the last one knew, do more with it, and pass a higher % forward.
W-Hats are just following trend
Collective intelligence has moved forward most with technology that reduces the friction of sharing knowledge: Language. Writing. Printing Press. Internet. Wifi. Mobile. We’ve advanced almost everything since cave people days. Our comms protocol (words) and bodies are rare exceptions. W-Hats address these two as part of the larger trend for the ultimate friction reduction.
We are cyborgs already
…via our smartphones, they just have very clumsy interfaces to brains. See the note about brain-machine bandwidth being the key.
Fin
Intrigued? Go put the real post in your queue!
—
Writing is a conscious effort to share thoughts. If you enjoyed this:
- Medium hearts make me smile☺
- FB and TW shares make me swoon!
If you want to see more of my writing you can follow: @a18n on Medium and Twitter.
— -
