Should LLMs drive cars? 🔥

How Ghost Autonomy burnt $220M trying to innovate in the self-driving industry.                                                                                                                                                                                                                                                                                                                                                                                                                 [Read Online]( The Startup That Wanted GPT to Drive Your Car How Ghost Autonomy burnt $220M trying to innovate in the self-driving industry. [Nicolás Cerdeira]( [fb]( [tw]( [in]( [email](mailto:?subject=Post%20from%20Failory&body=The%20Startup%20That%20Wanted%20GPT%20to%20Drive%20Your%20Car%3A%20How%20Ghost%20Autonomy%20burnt%20%24220M%20trying%20to%20innovate%20in%20the%20self-driving%20industry.%0A%0Ahttps%3A%2F%2Fnewsletter.failory.com%2Fp%2Fstartup-wanted-gpt-drive-car) Hey — It’s Nico. Another busy week in the startup bay. Ghost Autonomy shut down. Digital robot startup Autotab launched. Flip raised $144M; Varda, $90M. YC’s Winter 2024 Batch is out. All of this is brought to you by [Sidebar](, an exclusive leadership program designed to accelerate your career. Here’s what I got today: - The failure story of Ghost Autonomy, an autonomous driving startup that raised $220M 🔥 - An analysis of Autotab, the new AI that automates your tasks 📈 Presented by Sidebar Supercharge Your Career Ready to[supercharge your career]( in 2024? Introducing Sidebar,[an exclusive, highly curated leadership program]( designed to propel you to new professional heights. Picture yourself being a part of a dynamic network of peers, including Fortune 500 leaders and groundbreaking start-up CEOs, all offering their unique insights and expertise to fuel your journey. Imagine being able to tap into a[small group of supportive peers]( (think Fortune 500 execs and innovative start-up CEOs) who have best-in-class experience in their industry for advice, new perspectives, and raw feedback to catalyze your professional growth.[That’s Sidebar.]( The results are palpable. A staggering 93% of members affirm that Sidebar has been a game-changer in their professional path. Nothing propels your career trajectory like a robust peer group. Learn more at[Â]([sidebar.com](, and join thousands of top senior leaders from companies like Microsoft, Amazon, and Meta who have taken the first step towards accelerating their career. [Request Invite to Join →]( This Week In Startups 🫡 ICYMI How Grammarly writes emails [that users actually want to read](. 🔗 Resources The [step-by-step guide]( to building a $1M business. A failed startup founder shares her [top 10 learning points](. How this founder sold his business, and [why the process sucked](. The [YC 2024 Winter Batch]( is here — look at the 260 selected startups. Why [trying to find the perfect idea is bad]( for founders. What are VCs [really looking for right now](? 📰 News [Y Combinator looks for alumni investment]( for its upcoming fund. [A16z to invest $30M across Tech x Gaming startups]( via its Speedrun program. [Kauffman Fellows survey shows](that 53% of VCs plan to invest more in 2024. Cybersecurity software startup [Rubrik files for IPO](. 💸 Fundraising The shopping social network [Flip raised $144M](. [Varda raised $90M]( to help produce drugs in space. [SingleInterface raises $30M](to help small businesses get online. Fail(St)ory Ghost Crash It is no secret that the self-driving car sector has been facing challenges lately. Recently, we discussed [Phantom Auto’s shutdown](, a company working on remote-controlled cars. This week, Ghost Autonomy, a company backed by OpenAI, [announced it is shutting down]( globally. This news, [shared on its website]( just a few days ago, marks another major development in the industry. This comes half a year after Ghost Autonomy joined forces with OpenAI and got a $5 million investment, in addition to the [$220 million]( they had secured in prior funding rounds. Nevertheless, the CEO said that the company was shutting down because it was unable to continue funding itself and that the “path to profitability was uncertain.” Versatile Self-Driving From its inception in 2017, Ghost Autonomy had a clear mission-to develop an affordable, secure, and versatile autonomous driving kit for car manufacturers. Their vision was to create the Ghost Autonomy Engine, software that could lead to truly autonomous vehicles and potentially eliminate the need for driver supervision. The engine was intended to include all essential components of self-driving AI, complete with its own operating system and a software development kit (SDK) for manufacturers to integrate and tailor to their needs. Ghost Autonomy’s approach to autonomous driving was unique. Unlike most solutions that focus on object-detection neural networks, Ghost Autonomy developed a physics-based neural network. This network doesn’t identify objects, but models motion to prevent collisions. This method is supposed to be more cost-effective, faster to train, doesn’t require human data labeling, and demands less CPU power for in-car execution, making it accessible for more vehicles. A few months before its closure, Ghost Autonomy’s CEO, John Hayes, shared plans to enhance its self-driving technology by [incorporating LLMs](. The goal was to leverage the LLMs’ advanced reasoning to provide another layer of decision-making to its existing physics-based neural network. This integration aimed to help the AI handle complex scenarios, such as navigating through construction zones with temporary lanes or detours. However, some experts pointed out that this was not viable and that Hayes was just using the word LLM as a buzzword to get more funding. In the end, Ghost Autonomy did not succeed in integrating LLMs into their system. After exhausting the $220 million in funding they had accumulated, the company was unable to raise more capital and had to exit the market. Reasons for Failure Two weeks ago, when discussing Phantom Auto's shutdown, I shared this chart showing the [decline in funding]( in the autonomous driving industry since 2022. This data explains how Ghost Autonomy was able to raise so much money in its early days but failed to get new funding in the last year. The autonomous driving sector peaked in 2021 but has suffered a sharp decline. This has affected many startups in the industry that, like Ghost Autonomy, rely on this funding to survive. This situation highlights a broader issue: the lack of a short-term profitability roadmap for many startups in this field. Investing in autonomous driving is betting on a distant future, as widespread deployment of this technology remains a long way off. [Research from the Pew Research Center]( indicates that most Americans are still against the widespread use of autonomous driving. Moreover, manufacturers interested in adopting self-driving technology must navigate a complex web of regulations. There’s also the concern of associating their brand with any accidents that might occur due to the technology. Considering these factors, it’s understandable why many car manufacturers are cautious about investing in autonomous driving technology at this stage. This means that any startup developing self-driving AI will have a hard time reaching profitability in the near future and must find a way to fund itself in the short term. Go Deeper - Here is another article [analyzing the shutdown of Ghost Autonomy](. - This tech brief has all the information about [Ghost Autonomy’s technology](. Trend Radar Digital Robots At some point, we have all faced the monotony of tedious and repetitive tasks. Whether it's data entry, file organization, or synchronizing data across multiple databases, we’ve all pondered why, in 2024, there isn’t a more efficient method of handling these chores. Luckily, many new startups have emerged that promise to solve this problem. I am not talking about new automation or scraping tools, but rather about AI bots that understand the tasks you need to do and can complete them on their own. Just last week, the YC-backed digital robot [Autotab]( made its debut. [tw profile: Y Combinator] Y Combinator @ycombinator [tw] 🤖@autotabai (YC S23) is a digital robot that can operate a computer like a human. It handles your repetitive work and can run thousands of steps with 98%+ reliability. ycombinator.com/launches/Kmz-a… Congrats on the launch, @alexirobbins and @jonas_nelle! Apr 3, 2024 411 Likes 48 Retweets 13 Replies The idea is quite simple. You record the task you want to perform, and Autotab dissects it into actionable steps. Then, it operates discreetly in the background, liberating you to concentrate on more pressing matters. There is no need to program the bot. Autotab is designed to be as user-friendly as possible, requiring no coding experience. Founder Jonas Nelle [referred to Autotab]( as a “robot you can hire starting at $1/hour.” And it works non-stop, all day, every day. I am a bit skeptical about the accuracy of an AI like this one. In the past, I have dabbled into the world of scraping and web automation, and I know that getting the bot to do what you want is usually more complicated than it seems. I have also tried using tools like [Microsoft’s Power Automate](, which has recording functionalities that observe the user’s actions and attempt to deduce the steps behind them. However, I found this functionality to be not very precise and often required modifying a significant amount of code in the backend. Nevertheless, suppose Autotab truly can understand the user tasks with precision and can complete them with 98% reliability, as YC claims. In that case, I can absolutely see how a technology like this could revolutionize many industries. In fact, I cannot think of a single industry that would not benefit from technology like this. Every business has repetitive, mundane tasks like data entry. Autotab could significantly reduce the number of man-hours spent on these types of tasks, allowing for a substantial increase in productivity. Other Players - [Axiom]( is a no-code Chrome extension that employs browser bots to automate repetitive tasks. It is a powerful tool, but it is not as user-friendly as Autotab promises to be. There is no recording functionality, so you have to use their editor to configure the action workflow. - [Bardeen]( is similar to Autotab in the sense that they both try to leverage AI to make automation easier. However, Bardeen does so by making the user write what they wish to accomplish and then using LLMs to figure out the necessary steps. - [Microsoft’s Power Automate]( is really powerful. It can record user actions and autonomously create automation workflows. As I said, this function is not perfect and often requires tweaking the workflow manually. Nevertheless, as a robust no-code solution that leverages AI and comes pre-installed on all Windows computers, Power Automate is undeniably a strong competitor to Autotab. It is interesting that, although there are many other automation tools, none of them claim to be a digital robot available for hire on an hourly basis. This is a clever marketing strategy from Autotab: positioning their tool as something you hire rather than a product you purchase or subscribe to. The hourly pricing model further reinforces the concept that Autotab is not merely a tool but an additional member of the workforce. Refer Failory, Get Rewards Share Failory Chances are you have some more friends who would enjoy Failory as much as you do. Share Failory with these friends and cash in on premium resources and swag. You currently have 0 referrals, only 1 away from receiving my Pitch Deck Airtable. [Share Failory →]( Or copy and paste this link to others: Help Me Improve Failory How Was Today's Newsletter? If this issue was a startup, how would you rate it? [🚀 Launches to the moon!]( [🤔 Room for a pivot]( [💀 Crashes and burns]( That's all of this edition. Cheers, Nico Update your email preferences or unsubscribe [here]( © 2024 Failory 1309 Coffeen Avenue Ste 1200, Sheridan, Wyoming 82801, United States of America [[beehiiv logo]Powered by beehiiv](