Ulrich Gall

we didn't build flapping wings to fly why would we need consciousness for AI? evolution made things that contract we made things that rotate survival+reproduction had to be at the root of all action suffering, joy, love, hate, selfishness, altruism so much complexity we can skip all that and leave choices to the humans

@beffjezos @scottsullivan But… do we want to? Why would we? Seems like there are strictly simpler brains that are strictly more useful.

@scottsullivan my man I literally design brain-like computers for a living. Your brain is just a bunch of electrochemistry at the end of the day, and we can simulate it in stochastic electronics

The less people understand the flow of electrons through a computer, the more they will imagine fanciful thoughts like this. But no matter how many “if…then” statements or how many transistors, a computer will never become conscious just like my calculator will never be self-aware.

@pitdesi No prompt, btw… ! Just what grok thought would happen next.

@pitdesi I was curious what grok would do, result is actually really sweet. (After it refused twice… definitely pushing grok‘s comfort zone here, haha)

I always retweet when I see this picture of my wife on the timeline

@pitdesi And that was before AI, no?

Random idea: Fake window with monitor behind it. Camera tracks people and notices when someone looks at it. Then fast eye tracking to adjust the camera position on the AI world model (or 3D rendering) displayed in the fake window. Bonus points for doing this in a moving vehicle (boat?) and compensating for movement to provide fake stable horizon, might help with seasickness.

Dans le manifeste "techno-optimiste" de Marc Andreessen, il y a une phrase qui m'a marqué : "Our enemies are not bad people – but rather bad ideas." Nos ennemis ne sont pas des mauvaises personnes. Ce sont des mauvaises idées. Prenons Jancovici. L'homme est brillant, sincère, travailleur. Il ne se lève pas le matin en se disant qu'il va nuire à l'humanité. Mais l'idée qu'il porte la décroissance, le rationnement, la frugalité érigée en horizon civilisationnel est une idée profondément destructrice. Elle prend des esprits brillants et les transforme en commissaires politiques d'un futur appauvri. Et le plus fascinant, c'est ce que cette idée fait aux gens qui l'adoptent. Dans mon entourage, une grosse partie de mes amis est sur cette ligne décroissantiste, avec tout le package qui va avec. L'argent c'est mal mais ils en veulent. Il faut moins prendre l'avion mais ils rêvent de voyager partout. Il faut consommer moins mais ils ne renoncent à rien de ce qu'ils aiment vraiment. Et tous ont un point commun : ils sont déprimés. L'un d'eux m'a même confié qu'il était sous antidépresseurs. Ce n'est pas un hasard. C'est mécanique. Quand tu crois que ton désir de vivre, de créer, de t'élever est moralement suspect tu te détruis de l'intérieur. Tu passes ta vie à t'excuser d'exister. Tu vis dans la dissonance permanente entre ce que ton corps veut (plus, mieux, plus loin) et ce que ton idéologie t'ordonne (moins, sobre, immobile). D'où ma théorie : Quand on pense quelque chose de fondamentalement faux décroissance, communisme, extrémisme religieux (de tout ordre) ce n'est qu'une question de temps avant que ça devienne vraiment destructeur. D'abord pour soi. Puis pour les autres. Les mauvaises idées tuent. Lentement chez ceux qui y croient, brutalement chez ceux qui les subissent. C'est pour ça que la bataille des idées n'est pas un luxe d'intellectuel. C'est la bataille la plus importante de notre époque.

Here is some more technical detail. Also generated with AI (thanks, @perplexity_ai !) @tslaming nice job finding this and nice prompt for a smooth introduction. How do you get the formatting to look so good? Do you have a tool or do you edit manually? One thing that's interesting here is that this is continuation of an earlier patent that was filed just a few weeks earlier. Often, companies hold back detail so that year later they can later file a continuation to extend the lifetime of the patent. Cool that Tesla isn't doing that. Anyway - everything below, if it wasn't obvious, is AI generated. -- The Problem Being Solved The stated prior art problem is deceptively simple but has real manufacturing consequences at scale:[^4] Low-durometer (soft) hoses — cheap, conformable, seal well, but lack axial stiffness, so they deform, kink, or pull off barbs during installation. Robots struggle because the workpiece behaves unpredictably. High-durometer (hard/rigid) hoses — axially stiff and robot-friendly, but require very high push-on forces, risking barb damage or connector fracture. They also cannot accommodate barb-to-barb misalignment without leaking. Hose clamps — the traditional mitigation. They work, but each clamp is a separate BOM (bill of materials) part, adds weight, requires a torque step, consumes factory cycle time, and can be installed incorrectly. At the scale of a 4680 pack — which has many individual coolant barb connections distributed across the cooling distributers — even small per-connection times and failure rates compound severely.[^5][^6] The core tension: sealing robustness and geometric tolerance demand a soft inner interface; autonomous robot installation demands a stiff outer interface. Prior art forces a choice between the two. The Core Innovation: A Dual-Material Architecture Structural Concept The device is a short cylindrical sleeve, hollow throughout, constructed from two co-molded thermoplastic layers:[^3] Internal material: soft, rubber-type thermoplastic (nitrile, silicone, EPDM, fluorocarbon rubber, or similar elastomer), Shore A hardness roughly 5–90 depending on embodiment. This is the sealing layer — it deforms to fill surface asperities, minor scratches, dents, and debris on the barb. External material: rigid engineering thermoplastic — polypropylene (PP), polybutylene terephthalate (PBT), or nylon. Shore D hardness ≥50–70. This is the structural shell — it limits radial expansion of the soft inner layer, acts as a confinement ring that converts radial compression forces into hoop stress, and provides the axial stiffness needed for robotic push-on installation. The two materials are co-molded directly against each other with no adhesive or intermediate layer between them. Compatibility for direct bonding without delamination is therefore a material selection constraint, which is why the specific polymer families (thermoplastic elastomers against PP/PBT/nylon) are enumerated in the claims.[^3] The Flared End Geometry The geometry is where the mechanical elegance lies. Each end of the sleeve is flared — it opens outward — but the flare is not a simple linear taper. It is a three-segment non-monotone profile:[^3] Proximal portion (nearest the cylindrical body): slight flare angle relative to the axis Medial portion (mid-flare): steeper angle than the proximal portion Distal end (outermost lip): steeper still Formally, the internal angle increases from proximal → medial → distal. This creates a bell-mouth entry geometry that is progressively more open at the tip. During barb insertion: The barb tip contacts the distal (wide) entry, which acts as a lead-in funnel. Misalignment up to several centimeters is absorbed here.[^3] As the barb advances, the medial region begins to deflect inward and compress the soft inner layer against the barb's retaining ridge. At the proximal portion, the barb shoulder locks behind the barb ridge, the inner soft material is fully compressed in an annular sealing zone, and the rigid outer shell prevents the assembly from expanding radially — maximizing contact pressure for a given interference fit. The FEA simulations in the patent figures (4A vs. 4B) compare equivalent total strain in a single-material device vs. the two-material device. The two-material version shows significantly more uniform and controlled strain distribution — rather than stress concentrations at the barb ridge (which cause fatigue cracking in single-material elastomeric hoses), the stiff outer layer redistributes load around the circumference.[^3] Quantified Performance Claims These are not vague statements of improvement — the patent specifies measurable thresholds that can be independently tested: ParameterSpecified ThresholdSignificanceSeal integrity with barb surface defectsMaintains ≥207 kPa (30 psi) with debris/damage up to 250 µmCovers typical factory contamination from metal chips, coolant residue, coating imperfectionsSeal integrity with misaligned adjacent barbsMaintains ≥207 kPa even when adjacent barbs are misalignedCritical for 4680 pack assembly tolerancesMaximum burst toleranceUp to 689 kPa (100 psi) cited in embodiment ranges3× margin over nominal operating pressureNo additional attachment devicesNo hose clamps, zip ties, or crimp rings requiredDirect impact on BOM, weight, assembly cycle timeChemical resistanceEthylene glycol and automotive coolant compatibleLong-term pack serviceability [^3] A 250 µm debris tolerance is meaningful in context: typical automotive coolant circuit barbs have sealing surface roughness in the Ra 0.8–3.2 µm range, but powder coat drips, assembly debris, and minor handling damage commonly produce defects in the 50–500 µm range. The ≥207 kPa threshold at 250 µm debris means this connector is designed to be production-floor tolerant — rejects at final inspection are reduced by accepting barbs that would fail with a conventional hose-clamp assembly if debris were trapped under the clamp. Why This Matters for the 4680 Battery Pack Architecture Context: The Cooling Distributer Network The 4680 structural battery pack uses a tube-based cooling architecture where cooling distributers (manifold tubes that run lengthwise between cell rows) have barbs at regular intervals. These barbs connect to cross-tubes or direct coolant pathways that thermally interface with the cells. The number of such barb connections per pack scales with cell count — a 960-cell 4680 pack has a substantial number of individual hydraulic connections.[^7][^5][^6] In the prior 2170-cell architecture, hose clamps were used extensively. At the scale of Giga Texas/Berlin production rates, each manually torqued or robotically driven clamp represents cycle time, a possible torque error, and a potential leak point. The coupling device eliminates the clamp from the BOM entirely.[^4][^3] Autonomous Assembly Enablement The patent explicitly calls out "autonomous factory installation" as a design objective. This is not incidental. Tesla's Gigafactory automation strategy depends on moving from human-intensive clamp torquing to robot-pushed press-fit connections. For this to work reliably:[^4][^3] The push force must be within robot actuator limits (the stiff outer shell governs this — the device doesn't require excessive force because the inner material is soft enough to compress without catastrophic stress) The lead-in geometry must accommodate robot positional repeatability tolerances — a few mm of lateral misalignment during insertion cannot result in a damaged barb or a failed seal After installation, no secondary fastening step (clamp tightening, ferrule swaging) is needed — the device is self-retaining The three-segment flared entry directly addresses robot positional tolerance. Even a robot arm with ±2–3 mm TCP repeatability can successfully seat this device because the progressive funnel guides the barb tip into the sealing bore.[^3] Barb Geometry Compatibility Figure 3A of the patent shows three barb types in scope: double barb, single barb, and rounded single barb. The coupling device is designed to work across all three. This multi-barb-type compatibility is important because different subsystems (battery cooling distributers, motor coolant manifolds, inverter cooling ports) may use different barb geometries, and a single coupling device part number reduces supply chain complexity.[^3] What's Likely New in the 2026 Continuation The continuation (US 2026/0139776 A1, App. No. 19/443,561) was filed January 8, 2026 — mere weeks after the parent US 12,540,697 B2 was granted (February 3, 2026). In continuation practice, the applicant cannot add new matter, but can introduce new claims directed to aspects disclosed but not claimed in the parent. Based on the CPC classifications visible in the application header:[^8][^1] F28F 9/02 (2006.01): header box / end plate connections — this is a new classification vs. the parent, suggesting claims directed specifically to how the coupling attaches to a heat exchanger end plate or manifold header, not just a generic barb H01M 10/613, 10/625, 10/6568: same battery cooling classifications as parent F28F 9/02 is the key new CPC — the shift from F28F 9/0258 (quick-acting coupling) to F28F 9/02 (the broader header/end plate class) suggests the continuation may broaden or redirect claims toward how the assembled coupling device interfaces with the entire distributer/manifold structure, potentially covering the system-level integration (not just the sleeve geometry) more broadly The classification H01M 10/6568 (liquid coolant flow circuits for batteries) also points to possible method claims or apparatus claims covering the complete coolant loop architecture incorporating these couplings, which would be harder to design around than narrow device claims. Where This Works Best: Ideal Use Cases Optimal Deployment Scenarios 1. High-volume automated assembly of battery packs Any application where robotic arms push-connect coolant hoses without human verification. The self-guiding bell-mouth entry and self-retaining press-fit eliminate inspection steps and torque verification.[^4][^3] 2. Dense barb arrays with positional tolerance stack-up In the 4680 pack, cooling distributers running lengthwise create many barb connection points across the width of the pack. Manufacturing tolerances in distributer placement, cell stack height, and barb spacing can compound into several mm of positional error. The misalignment tolerance (stated as up to multiple centimeters for adjacent barbs) directly addresses this.[^3] 3. Environments with metal chip / debris contamination Machined aluminum barbs and powder-coated distributers shed particulate. A ≤250 µm debris-tolerant seal eliminates the need for clean-room-level cleanliness at the barb connection interface.[^3] 4. Weight-sensitive and BOM-sensitive platforms Eliminating hose clamps reduces part count and mass. At EV pack scale this compounds across dozens or hundreds of connections. 5. Ethylene glycol coolant circuits, standard operating temperatures and pressures The device is chemically compatible with EG/water mixes. At 30 psi (207 kPa) operating pressure, it works for typical liquid-cooled battery thermal management circuits which operate at 15–30 psi.[^3] Where It Is Less Applicable High-pressure hydraulic circuits (>100 psi / 689 kPa sustained): the device is not intended as a hydraulic pressure-line fitting; O-ring face seals or swaged fittings remain appropriate there. High-temperature coolant paths (glycol > ~120°C sustained): melt temperature limits of the internal thermoplastic impose an upper bound; powertrain loops near turbochargers or exhaust-adjacent components may exceed this.[^3] Applications requiring field re-mate cycles: the device is designed for one-time press-on installation in a factory. It is not a quick-disconnect; serviceability in the field requires cutting the coupling off and replacing it, which is practical for a pack-level replacement but not a frequent reconnect scenario. Non-barb interfaces: the design specifically targets male barb connectors. It does not apply to threaded fittings, flange joints, or face-seal interfaces. Prior Art Differentiation The key distinguishing axis vs. the cited prior art: Prior Art ReferenceWhat it doesWhat Tesla's device does differentlyUS 2005/199308 (Parker push-on hose)Single-layer push-on hose with low-friction coreTwo distinct hardness layers; flared geometry not present; requires sealantFR 3080166 (Coutier fluid dispenser ramp)Fluid manifold ramp designNot a coupling device per se; no dual-material press-fit sleeveUS 10,385,999 (Terumo CV fluid coupling)Medical fluid circuit couplingNot designed for robotic automotive assembly; different operating pressure/chemical regimeTraditional hose + clampClamp creates mechanical retentionNo clamp required; self-retaining; no secondary fastening stepHigh-durometer rigid hoseAxially stiff, robot-installableBut requires high insertion force and cannot tolerate misalignment without leaking [^1][^3] Material Engineering Considerations The internal/external material pairing is constrained by several simultaneous requirements:[^3] Direct bonding compatibility — the two polymers must adhere to each other during co-molding without delamination under thermal cycling (automotive battery packs see −40°C to +70°C range in service) Differential thermal expansion — if the CTE (coefficient of thermal expansion) of the outer material is significantly higher than the inner, thermal cycling could loosen the bond. PP has CTE ~150 ppm/°C; rubber elastomers typically ~200 ppm/°C — a modest mismatch that the geometry accommodates Chemical compatibility — both layers must resist EG/water coolants. EPDM and nitrile rubber have good EG resistance; PP and PBT both resist EG well Melt temperature ordering — the internal material must have a lower melt temperature than the external material, which governs co-molding process: the outer shell is formed first at higher temperature; the inner layer is overmolded or co-extruded at lower temperature without remelting the shell This is essentially a two-shot injection molding or co-extrusion process that is well-established in the rubber/thermoplastic composite industry, which means Tesla is not inventing a new manufacturing process — they are applying dual-material molding to a geometric form factor and application (automotive coolant barb coupling) that hadn't used it this way before. Manufacturing and Production Economics At Giga Texas/Berlin production rates, the cost impact of eliminating hose clamps from a battery pack coolant network is non-trivial:[^5][^6] Part count reduction: a hose clamp is a separate part with its own supply chain, quality inspection, and installation tooling Cycle time: a robotic clamp-tightening step takes several seconds per connection; a press-fit takes under one second Yield improvement: clamp torque errors (under- or over-torque) are a common source of coolant leaks at end-of-line. The press-fit has no torque variable — it either seats fully or it doesn't, making pass/fail detection simpler Rework reduction: barbs with minor surface defects that would previously require manual clamp adjustment or rework can now be sealed by the device's debris tolerance These are incremental gains per connection, but at EV pack manufacturing scale, they represent meaningful cost and quality improvements consistent with Tesla's Design for Manufacturing philosophy as seen across the 4680 program.[^7][^6]

@GeniusGTX Arbitrage opportunity: I would think the significant others are capable, smart people, too. So - start a location independent business (that benefits from an on-site team) in Brownsville? Or a startup incubator that also offers investment, with some funding from SpaceX?

Elon Musk says one constraint blocks every Starbase hire: Their partner's job. Brownsville, Texas. Population 187,000. The nearest tech hub: Austin. 350 miles north. Engineers had to relocate everything. Houses sold. Schools researched. Career ladders restarted. "Getting engineers to move… I call it the significant other problem." A spouse with a career. Kids in school. A partner's job market. The technical work was the easy part of recruiting. The recruiting problem started somewhere else. Then Musk explained why Brownsville was a harder sell than Silicon Valley. "For Starbase that was particularly difficult, since the odds of finding a non-SpaceX job… are pretty low." He named the constraint: **the significant other problem**. A name the recruiters used, the engineers admitted, and the partners felt first. Musk, who had moved his own household to Texas for SpaceX, had paid the same tax. And kept Tesla majority in California, where the talent could still spouse-shop. A startup hire at a Silicon Valley campus could promise their spouse a dozen recruiters within driving distance, while a Starbase hire could promise rocketry, palm trees, and almost nothing else by way of secondary employment. "It's quite difficult. It's like a technology monastery thing, remote and mostly dudes." One household. Two careers. One Brownsville. Two careers and one zip code rarely worked out. After Musk named the problem, the recruiting funnel narrowed to the single, retired, and remote-spouse. Musk, on what the technology monastery cost: "Not much of an improvement over SF." What's the constraint on hiring at your company that has nothing to do with the work? P.S. I made a full playbook breaking down the timeless decision-making mental models used by history's greatest thinkers. Comment "models" and follow @GeniusGTX so I can DM you a copy. — Elon Musk ( @elonmusk ), CEO of Tesla and SpaceX, on Dwarkesh Patel's ( @dwarkesh_sp ) podcast

@TonyTaylor_87 This is my video. He had fairly recently had neck / upper back surgery and wasn’t feeling the best during this interview. The nearly two hour interview would’ve gone on longer if he wasn’t in pain

So sorry to hear!! Much easier would be to get some blood tests done. When you're feeling ok, figure out which ones would be useful, and then arrange with a friend to call a blood draw service to your house at your request. Then it's one simple message, minimal hill climb required.

I had a couple hours of severe depression today where I was hardly able to move. It’s been a while since I felt that. It’s not the same as being “just tired” For the first time, I wonder if I could try to get a brain scan during such a moment? But these moments are unpredictable

@Mayer It would be SO FUNNY if Boca Chica, Chiriquí happened to be a good location for this. Sadly, it's not.

Panama is only 9° from the equator. 🚀 That means Starship gets more free velocity from Earth’s rotation on every eastward launch = more payload, less fuel, lower costs. No hurricanes. Two-ocean access for recovering rockets. Physics already did half the pitch for us. 🇵🇦

@Mayer @AMOTTAPANAMA Yes!!!

No. They should stay “greedy” and invest in the most profitable way possible and get a kick out of seeing the “net worth” grow and grow. Because that means that the fruits of their hard work are allocated most efficiently for the benefit of all, maximizing the “net worth” they had for society - their legacy.

When it comes to billionaires my take is the opposite of Leftists: Billionaires SHOULD consume all their wealth: megayachts, English castles, mountains of cocaine, own a dozen homes and 30 cars - do it! Spend it all! The worst thing that can happen is for a billionaire to try to ‘do good in the world’ and shovel money to NGOs for some cause they get excited about. That’s the worst thing that can happen. Charitable donations should be taxed at 500%. A billionaire’s lavish lifestyle poses no threat to me but their misguided ‘good intent’ is absolutely cancerous and potentially civilization ending.

What subsidies? The $465M DOE loan Tesla repaid early? Regulatory credits other carmakers pay Tesla for failing emissions rules? Or NASA paying SpaceX for actual launches that saved the US space program? GM got $50B+ bailout (never fully earned back). Legacy auto: trillions in protection + subsidies for decades of mediocre cars. Elon: took the same game, repaid what he borrowed, and delivered reusable rockets + EVs that actually work. Ah sorry I just saw you have a parody account so you're just making fun of people who actually think the nonsense you posted there makes any sense?

Says the guy whose companies were built with billions in government subsidies, tax credits, NASA contracts, federal loans, carbon credits, and public infrastructure. Nothing screams “free market capitalism” like asking taxpayers to finance your empire while calling everyone else socialist.

We must preach the gospel of free market capitalism to all who will listen. It is the answer to our problems.

Space launch was a clear case where there was a large difference in efficiency between what was possible and what was done in practice before SpaceX. A large part of that was due to everything being locked in to what (just barely) already worked, with huge risk aversion. WIth national prestige or a half billion dollar geosync satellite on the line, speculative engineering ideas that might result in a public debacle were not welcome. When failure is not an option, success can stay very expensive. You need to experiment to improve, and that fundamentally means being comfortable with failure. If you know it is going to work, it isn’t an experiment. I have long believed that nuclear power today is in precisely the same state as space launch two decades ago, but the even more pressing question now is if semiconductor fabrication might also be. On the one hand, Moore’s Law has been a sequence of heroic miracles of technology at the wafer fabrication level, grinding out hundreds of compounding small improvements. On the other hand, fabs are “too big to fail”, and there are elements of extreme conservatism at play. Intel’s “Copy exactly!” fab development exemplifies that mindset – instead of every new building being an opportunity to explore and optimize processes, it was deemed more valuable to just replicate. While each individual machine may be straining against physical limits of technology, it is possible that the systems orchestrating them all together could be far from optimal. The explore / exploit axis is fundamental to all decision making, but human risk avoidance probably biases away from optimal exploration.

The Copa app showed the tickets correctly and let me buy seats, and during seat selection it correctly showed two adults and one infant. So - between that and Copa being able to fix it in the end, it looks like a Copa issue to me.

Hey @CopaAirlines and @Expedia , y'all got some issues to work out. I bought tickets for two adults and one (in-seat) infant on Expedia, and somehow Copa thought the tickets were for 3 adults. This was discovered at Check-In, and we almost missed our flight. The gate agent told me to call Copa customer service or Expedia (and said they can't do it themselves), or buy another ticket (!). I called both, neither of them were able to resolve it, and in the end the check-in agent's supervisor was able to fix it. In the middle of it some other traveler walked by and said she had the same issue - and no infants involved in her case.

Und wieder einmal ist es soweit: Der ÖRR mit einem weiteren "Hit Piece" über Tesla und Elon Musk 🤦‍♂️ "ELON MUSK UNCOVERED - Das Tesla-Experiment", so die neueste "Doku" bei der ARD. @OERRBlog Schon lustig irgendwie: Die Leute verlassen X, weil angeblich ach so böse Desinformation dort verbreitet wird, aber man muss nur ARD einschalten und bekommt dort wirklich Propaganda vom Allerfeinsten serviert, finanziert von unseren Gebühren, im direkten Widerspruch zu Programmauftrag und Medienstaatsvertrag. Aber das stört anscheinend niemanden. Vorab: Es gibt in dem gesamten Bericht keinerlei belastbaren Zahlen, keine Statistiken, keine Bewertungen, keinerlei Einordnungen von Informationen - sondern von vorne bis hinten nur eines: Emotionen und Narrative. Als Ausgangspunkt für die "Dokumentation" hat man wieder mal die Handelsblatt-"Recherche" zu den so genannten "Tesla-Files" als Anlass genommen und daraus das folgende Narrativ gebastelt - bitte festhalten: Skrupelloser Silicon-Valley-Milliardär, der die Menschen ohnehin nur als Übergangsphänomen sieht auf dem Weg hin zu einer irgendwie gearteten post-humanistischen Utopie, in der Technologie "die menschliche Spezies überwindet", geht für das Erreichen seiner Ziele über Leichen und setzt mit seiner gefährlichen Autopilot-Technologie Menschenleben aufs Spiel und missachtet dabei natürlich Regeln und Gesetze. So ungefähr. Eingerahmt wird der Bericht am Anfang und Ende von dem tragischen Unfall eines Familienvaters, der mit seinem Tesla tödlich verunglückt ist - emotional suggerierend, dass es irgendwie der Tesla gewesen sei, der den Unfall verursacht bzw. zu verschulden habe (ohne natürlich irgendwelche konkreten Beweise oder Ergebnisse dazu zu benennen). Selbst wenn es diese Beweise gäbe, dann müsste man als nächstes einordnen, inwiefern das Führen eines Teslas in irgendeiner Weise tatsächlich gefährlicher sein soll als das Führen jedes anderen Fahrzeugs auf der Welt, auch oder gerade mit Verwendung von Fahrassistenzsystemen. Bewusst wird auch keine Unterscheidung in dem Bericht vorgenommen zwischen dem System "Autopilot" und dem so genannten autonomen Fahren, an dem Tesla mit seinem FSD-System seit vielen Jahren arbeitet. Man kann ja durchaus kritisieren, dass Tesla bzw. Elon Musk mit seinen Aussagen zu "jeder Tesla hat seit Jahren einen FSD-Computer an Bord und wird bald autonom fahren können (Potenzial für vollautonomes Fahren)" falsche Hoffnungen geweckt hat, insbesondere was den Zeitstrahl dieser Entwicklung angeht, aber mit keinem Wort wird erwähnt, dass jede Selbstfahr-Technologie strengen Sicherheitsauflagen unterliegt, Untersuchungen der NHTSA in den USA bei Unfällen nichts ungewöhnliches sind (und bislang auch noch zu keinen negativen Ergebnissen führten), und die Einführung von Full-Self-Driving insbesondere in Europa aufwändigsten und strengsten Kontrollen unterliegt, siehe jetzt den Bericht der niederländischen RDW. Dass Teslas Fahrzeuge durch objektive Sicherheitstests und Prüfbehörden feststellbar zu den sichersten Fahrzeugen der Welt gehören, das darf in einem solchen Bericht natürlich nicht erwähnt werden, stattdessen werden einzelne fatale Unfälle aufgeführt, um irgendwie den Eindruck zu erwecken, dass Teslas irgendwie unsicher seien und jeder, der einen Tesla kaufe, gefühlt jederzeit Gefahr laufe, dass der Tesla ihn als nächstes gegen einen Baum fahre - und der Fahrer dann anschließend aufgrund nicht öffnender Türen im Fahrzeug verbrenne. Neben altbekannter Anti-Tesla-Prominenz wie Edward Niedermeyer, die selbstverständlich ausführlichst in der Doku zu Wort kommen dürfen, darf natürlich auch eine Pseudo-Einordnung irgendeines Psycho-Gurus nicht fehlen, der uns erklärt, was die eigentlich Philosophie hinter diesen ganzen Silicon-Valley-Oligarchen so sein soll. Die lautet ungefähr so: Musk (und all die anderen) arbeiten als so genannte Longtermists an einer gefühlt menschenfeindichen Utopie, in der Technologie quasi über dem Menschen steht, und um dieses Ziel zu erreichen, sind alle Mittel Recht, weil das Große-Ganze ja die Utopie ist und so. Dass das alles diametral entgegengesetzt zu Musks Aussagen und Taten steht, Musk ja selbst einer der größten Warner vor den Gefahren von Technologien wie Künstlicher Intelligenz ist, dafür OpenAI mitgegründet hat, damit diese Technologie eben nicht profitorientiert und möglicherweise menschenfeindlich eingesetzt werden kann, und auch seine Philosophie bezüglich Mensch vs. Technologie ja eine andere ist (siehe seine Diskussionen mit z.B. Sam Altman) - das alles wird natürlich überhaupt nicht erwähnt. Stattdessen wird sein Unternehmen Neuralink nur kurz in dem Kontext erwähnt, dass der Eindruck entsteht, Neuralink wäre genau dazu da, um die menschliche Spezies quasi durch den Upload von Gehirnen in die Cloud überflüssig zu machen. Nicht fehlen darf natürlich auch die Erwähnung der politischen Unterstützung Trumps, die angeblich nur deshalb erfolgt, um sich Trump zu erkaufen und sich so von den Untersuchungen zu all den angeglichen halblegalen oder illegalen Machenschaften Musks freizukaufen. Das alles ist so dermaßen einseitig und so absurd verfälscht, dass man gar nicht weiß, wo man da anfangen soll, das wieder halbwegs gerade zu biegen. Niemand, der nicht die Hintergründe und Informationen hat, kann bei solchen angeblichen Dokumentationen durchschauen, wie er von vorne bis hinten emotional über den Tisch gezogen wird. Es gibt viele Dinge, die man ja wirklich bei Tesla und bei Elon Musk kritisieren kann, natürlich gibt es die. Aber solche Dokumentation erfüllen absichtlich einen ganz anderen Zweck - leicht zu erkennen für jeden, der die Hintergründe kennt. Erschreckend ist, dass das ja nicht irgendjemand privat angefertigt hat, was ja noch halbwegs erklärbar wäre, sondern dass so etwas tatsächlich (und immer und immer wieder) im öffentlich-rechtlichen Fernsehen, welches eigentlich zu Neutralität und Objektivität verpflichtet ist, weggesendet wird. ardmediathek.de/video/elon-mus…

@IceSolst I agree reading compiler output is silly but people do read the output of human software engineers. And that is the analogy that matters.

Interesting article on treating agent output like compiler output (and why) skiplabs.io/blog/codegen_a…