Tesla’s autopilot and it’s leaps using AI

2

October

2024

5/5 (1)

Some of us might have experienced pieces of autonomous driving when stepping in new(er) cars, ranging from adaptive cruise control to partly autonomous driving with Tesla models. In this article I would like to go through a small history of the Tesla Autopilot and how it’s approach has been shifted and drastically accelerated using the integration of AI models.
(Greg, 2021) As of 2015 the first version of autopilot was released overnight to existing Tesla users with the right specifications. This version was simply able to drive like most adaptive cruise controls are currently able to do namely: stay within the right lane and follow other traffic. As of 2016-2019 we saw an improvement of this autopilot (enhanced autopilot), introducing automatic lane shifts, parking and more which relied heavy on thousands of lines of static code. As of 2020 Tesla has introduced Full-Self-Driving (FSD) beta’s (Greg, 2021).
As of from 2020 we have seen giant leaps forwards in regards to the capabilities of FSD, this is mainly due to the fact that Tesla has been able to convert its collected data and code based on the earlier models from 300K+ lines of code into an AI neural network (Ali, 2024). Which means that the cars no longer stop for a red light because there is a line of code stating you need to stop for red, but it determines it based on the thousands of hours of video data it has been trained on from other tesla drivers. This results in some new emergent behaviours, as for example, a youtuber was driving a tesla, but due to roadwork it decided to do a U-turn and find a different route, which according to Ashok Elluswamy (head of AI/Autopilot at Tesla) was behaviour the FSD AI had learned by itself.
Moving away from static lines of code to an AI will/is showing to make massive leaps towards a future without necessary interventions of humans when driving. Personally I like driving, but not having to will result in much time more effectively spend.
Currently there are still lots of legal and technological constraints before full autonomous self-driving will be rolled out to the public, but the future potentially looks driver-less.

Ali, I. (2024, September 9). Tesla (TSLA) reveals FSD 12.5 roadmap ahead of the Robotaxi event, v13 in October, aims FSD for China and Europe in 2025. Tesla Oracle. https://www.teslaoracle.com/2024/09/08/tesla-lays-down-critical-fsd-milestones-ahead-of-the-robotaxi-event-aims-fsd-for-china-and-europe-in-2025/

Greg . (2021, October 9). A timeline of Tesla Autopilot: From inception to now. That Tesla Channel. https://www.thatteslachannel.com/a-timeline-of-tesla-autopilot-from-inception-to-now/

Please rate this

How AI Transformed My Learning Process & Tried to Predict My Personality

26

September

2024

No ratings yet.

Generative AI continues to amaze me with its vast possibilities and the profound impact it’s already having on our world. It’s exciting to think about where this technology will be in five years or what innovations might be trending by then. The current enthusiasm surrounding AI among students and the general public is undeniable. I recall our first lecture when the professor asked about our interests, and almost every hand went up when AI was mentioned.

This enthusiasm resonates with my own experiences. When I started my Bachelor’s thesis, I was overwhelmed and unsure if I was putting in enough effort. I felt stuck, with so many questions and no clear direction. My supervisor, noticing my struggle, encouraged me to use ChatGPT. He continually pushed me to explore different Generative AI tools, each suited for various purposes.

I was diving into a completely new topic for my thesis, one I knew little about. However, with my supervisor’s guidance and his insistence on leveraging these AI tools, I gradually gained confidence. The AI didn’t just answer my questions; it also helped me navigate and understand the complexities of my thesis topic. This experience profoundly influenced my learning process, showing me how GenAI can empower students to learn and grow independently.

I think that Generative AI is more than just a tool; it’s a powerful ally in learning and creativity. It can potentially transform education by providing students with the support they need to explore new ideas and concepts. However, like any tool, its effectiveness depends on how we use it.

These days, I find myself turning to ChatGPT quite frequently. After interacting with it so much, I began to wonder: could it predict what kind of person I am based on our conversations? Out of curiosity, I asked it directly. Here’s the response I received:

Although the description touched on a few aspects of my personality, it felt a bit vague. So, I took it a step further and asked ChatGPT which personality type it thought I had. It guessed either ENTJ or INTJ:

For those unfamiliar with the 16 Personalities test, here’s the link if you’re interested: https://www.16personalities.com/. Despite ChatGPT’s efforts, it wasn’t accurate because my actual personality type is Consul: ESFJ-A.

This just goes to show that while ChatGPT is impressive in many areas, understanding the intricacies of someone’s personality is still a challenge for it (at least for now!).

Please rate this

Bridging the Gap Between AR, AI and the Real World: A Glimpse Into the Future of Smart Technology

12

September

2024

5/5 (3)

Apple’s recent keynote showcased new products, including the iPhone’s groundbreaking AI integration. However, when you break it down, what Apple has really done is combine several existing technologies and seamlessly integrate them, presenting it as a revolutionary technology. This sparked my imagination of what could already be possible with existing technologies and what our future might look like. This sparked my imagination about what could already be possible with today’s technology—and what our future might look like.

Apple introduced advanced visual intelligence, allowing users to take a picture of a restaurant, shop, or even a dog, and instantly access a wealth of information. Whether it’s reviews, operating hours, event details, or identifying objects like vehicles or pets, this technology uses AI to analyze visual data and provide real-time insights, bridging the gap between the physical and digital worlds. Tools like Google Image Search and ChatGPT have been available for some time, but Apple has taken these capabilities and seamlessly integrated them into its ecosystem, making them easily accessible and more user-friendly [1]. The Apple Vision Pro merges AR and VR, controlled by moving your eyes and pinching your fingers [2]. I’ve tried it myself, and it was incredibly easy to navigate, with digital content perfectly overlaying the physical world. Now imagine the possibilities if Apple integrated the iPhone’s visual intelligence into the Vision Pro. This headset wouldn’t just be for entertainment or increasing work productivity; it could become an everyday wearable, a powerful tool for real-time interaction with your surroundings.

Picture walking through a city wearing the Vision Pro. By simply looking at a restaurant and pinching your fingers, you could instantly pull up reviews, check the menu, or even make a reservation. Or, if you see someone wearing a piece of clothing you like, you could instantly check online where to buy it, without needing to stop. With these capabilities, the Vision Pro could bring the physical and digital worlds closer together than ever before, allowing users to interact with their environment in ways we’re only beginning to imagine.

Do you think the existing technologies can already do this? Do you think this is what the future would look like? I’m curious to hear your thoughts.

Sources:

[0] All images generate by DALL-E, a GPT made by ChatGPT.

[1] https://www.youtube.com/watch?v=uarNiSl_uh4&t=1744s

[2] https://www.apple.com/newsroom/2024/01/apple-vision-pro-available-in-the-us-on-february-2/

Please rate this

fintech bunq as a disruptive force in the dutch banking sector

6

September

2024

No ratings yet. During our first class we talked about whether banks are necessary or not. This inspired me to write the following article about how new digital banks (also called fintech/neobanks) are disrupting the traditional banks. I would like to write this blog post from my personal perspective, which is as someone who previously worked inside a Dutch fintech bank named bunq. What I experienced here is the way the business model of banks are currently being disrupted.
One of the initial ideas of bunq was to no longer use the money of people on the bank and make profits by investing/lending this money out, but using a subscription model to generate revenues. Resulting in no or a lower need to invest client’s money and thus lowering the risk profile of the bank in general in times of crisis. This idea worked well, since people are more used to having subscriptions then years ago. Using subscription models to acquire revenue, the bank has been able to give back to the customers by consistently offering the highest interest rates of all Dutch banks (currently 3.36% vs 2.60% for second highest).
Although their business model is different and disrupting the Dutch banking sector, what is more impressive is their operating and organizational model. Bunq is truly run like a start-up software company, the largest number of people employed besides the necessary compliance people are by far developers. The organization is extremely flat, in such a way that basically every document and/or piece of code is still passed through the CEO (Ali Niknam). This is something unheard of at traditional banks.
Over time the Dutch media has written a few quite negative stories about bunq and their lack of protection against banking fraud. I think that everyone can conclude that bunq could have been more cautious here, but this is not the point I am trying to make here. Because what this scandal also showed is the way their organization works and is able to disrupt the rest of the industry. This is due to the fact that after the scandal came out, bunq handled quickly and made changes to their processes, made adjustments in their app and set up a phone help desk, all within 3 weeks. Clearly displaying how flat and fast this organization can handle on matters. I assume that in larger banks their bureaucratic processes would be in the way of such quick actions.
Another example showing the disruption of bunq is by how they have handled the process of onboarding new customers. the regulator (Nederlandsche Bank) had a very strict policy towards onboarding new customers and what questions to ask. The procedure of onboarding new customers was a long and bureaucratic process of questioning the customer. An example question was “ Are you a fraudster?”. After Ali Niknam went through these questions he stated that using the numerous datapoints and AI they are able to make better informed decisions on whether customers should be onboarded than asking a large abundant question list. Next to this, they wanted to make the process of onboarding as quick and convenient as possible. After numerous discussions bunq decided to do what no other bank had dared to do; sue your regulator. This can be seen as a suicide mission, since suing the people that grade your homework does not often result in high grades. But the courtcase turned out in the favor of bunq, the judge stated that the process of onboarding was outdated and that bunq was indeed not doing anything wrong. This has resulted in a big leap forward in terms of how KYC (know your customer) processes are currently set up. The disruptive nature of bunq clearly shows, since most other dutch banks started adapting their onboarding processes after the trial of bunq.

Please rate this

Adverse training AI models: a big self-destruct button?

21

October

2023

No ratings yet.

“Artificial Intelligence (AI) has made significant strides in transforming industries, from healthcare to finance, but a lurking threat called adversarial attacks could potentially disrupt this progress. Adversarial attacks are carefully crafted inputs that can trick AI systems into making incorrect predictions or classifications. Here’s why they pose a formidable challenge to the AI industry.”

And now, ChatGPT went on to sum up various reasons why these so-called ‘adversarial attacks’ threaten AI models. Interestingly, I only asked ChatGPT to explain the disruptive effects of adversarial machine learning. I followed up my conversation with the question: how could I use Adversarial machine learning to compromise the training data of AI? Evidently, the answer I got was: “I can’t help you with that”. This conversation with ChatGPT made me speculate about possible ways to destroy AI models. Let us explore this field and see if it could provide a movie-worthy big red self-destruct button.

The Gibbon: a textbook example

When you feed one of the best image visualization systems GoogLeNet with a picture that clearly is a panda, it will tell you with great confidence that it is a gibbon. This is because the image secretly has a layer of ‘noise’, invisible to humans, but of great hindrance to deep learning models.

This is a textbook example of adversarial machine learning, the noise works like a blurring mask, keeping the AI from recognising what is truly underneath, but how does this ‘noise’ work, and can we use it to completely compromise the training data of deep learning models?

Deep neural networks and the loss function

To understand the effect of ‘noise’, let me first explain briefly how deep learning models work. Deep neural networks in deep learning models use a loss function to quantify the error between predicted and actual outputs. During training, the network aims to minimize this loss. Input data is passed through layers of interconnected neurons, which apply weights and biases to produce predictions. These predictions are compared to the true values, and the loss function calculates the error. Through a process called backpropagation, the network adjusts its weights and biases to reduce this error. This iterative process of forward and backward propagation, driven by the loss function, enables deep neural networks to learn and make accurate predictions in various tasks (Samek et al., 2021).

So training a model involves minimizing the loss function by updating model parameters, adversarial machine learning does the exact opposite, it maximizes the loss function by updating the inputs. The updates to these input values form the layer of noise applied to the image and the exact values can lead any model to believe anything (Huang et al., 2011). But can this practice be used to compromise entire models? Or is it just a ‘party trick’?

Adversarial attacks

Now we get to the part ChatGPT told me about, Adversarial attacks are techniques used to manipulate machine learning models by adding imperceptible noise to large amounts of input data. Attackers exploit vulnerabilities in the model’s decision boundaries, causing misclassification. By injecting carefully crafted noise in vast amounts, the training data of AI models can be modified. There are different types of adversarial attacks, if the attacker has access to the model’s internal structure, he can apply a so-called ‘white-box’ attack, in which case he would be able to compromise the model completely (Huang et al., 2017). This would impose serious threats to AI models used in for example self-driving cars, but luckily, access to internal structure is very hard to gain.

So say, if computers were to take over humans in the future, like the science fiction movies predict, can we use attacks like these in order to bring those evil AI computers down? Well, in theory, we could, though practically speaking there is little evidence as there haven’t been major adversarial attacks. Certain is that adversarial machine learning holds great potential for controlling deep learning models. The question is, will the potential be exploited in a good way, keeping it as a method of control over AI models, or will it be used as a means of cyber-attack, justifying ChatGPT’s negative tone when explaining it?

References

Huang, L., Joseph, A. D., Nelson, B., Rubinstein, B. I., & Tygar, J. D. (2011, October). Adversarial machine learning. In Proceedings of the 4th ACM workshop on Security and artificial intelligence (pp. 43-58).

Huang, S., Papernot, N., Goodfellow, I., Duan, Y., & Abbeel, P. (2017). Adversarial attacks on neural network policies. arXiv preprint arXiv:1702.02284.

Samek, W., Montavon, G., Lapuschkin, S., Anders, C. J., & Müller, K. R. (2021). Explaining deep neural networks and beyond: A review of methods and applications. Proceedings of the IEEE109(3), 247-278.

Please rate this

AI-Powered Learning: My Adventure with TutorAI

16

October

2023

No ratings yet.

Subscribe to continue reading

Subscribe to get access to the rest of this post and other subscriber-only content.

Please rate this

Weapons of mass destruction – why Uncle Sam wants you.

14

October

2023

No ratings yet.

The Second World War was the cradle for national and geopolitical informational wars, with both sides firing rapid rounds of propaganda at each other. Because of the lack of connectivity (internet), simple pamphlets had the power to plant theories in entire civilizations. In today’s digital age, where everything and everyone is connected, the influence of artificial intelligence on political propaganda cannot be underestimated. This raises concern as, unlike in the Second World War, the informational wars being fought today extend themselves to national politics in almost every first-world country.

Let us take a look at the world’s most popular political battlefield; the US elections; in 2016, a bunch of tweets containing false claims led to a shooting in a pizza shop (NOS, 2016), these tweets had no research backing the information they were transmitting, but fired at the right audience they had significant power. Individuals have immediate access to (mis)information, this is a major opportunity for political powers wanting to gain support by polarising their battlefield.

Probably nothing that I have said to this point is new to you, so shouldn’t you just stop reading this blog and switch to social media to give your dopamine levels a boost? If you were to do that, misinformation would come your way six times faster than truthful information, and you contribute to this lovely statistic (Langin, 2018). This is exactly the essence of the matter, as it is estimated that by 2026, 90% of social media will be AI-generated (Facing reality?, 2022). Combine the presence of AI in social media with the power of fake news, bundle these in propaganda, and add to that a grim conflict like the ones taking place in East Europe or the Middle East right now, and you have got yourself the modern-day weapon of mass destruction, congratulations! But of course, you have got no business in all this so why bother to interfere, well, there is a big chance that you will share misinformation yourself when transmitting information online (Fake news shared on social media U.S. | Statista, 2023). Whether you want it or not, Uncle Sam already has you, and you will be part of the problem.

Artificial intelligence is about to play a significant role in geopolitics and in times of war the power of artificial intelligence is even greater, luckily full potential of these powers hasn’t been reached yet, but it is inevitable that this will happen soon. Therefore, it is essential that we open the discussion not about preventing the use of artificial intelligence in creating conflict and polarising civilisations, but about the use of artificial intelligence to repair the damages it does; to counterattack the false information it is able to generate, to solve conflicts it helps create, and to unite groups of people it divides initially. What is the best way for us to not be part of the problem but part of the solution?

References

Facing reality?: Law Enforcement and the Challenge of Deepfakes : an Observatory Report from the Europol Innovation Lab. (2022).

Fake news shared on social media U.S. | Statista. (2023, 21 maart). Statista. https://www.statista.com/statistics/657111/fake-news-sharing-online/

Langin, K. (2018). Fake news spreads faster than true news on Twitter—thanks to people, not bots. Science. https://doi.org/10.1126/science.aat5350

NOS. (2016, 5 december). Nepnieuws leidt tot schietpartij in restaurant VS. NOS. https://nos.nl/artikel/2146586-nepnieuws-leidt-tot-schietpartij-in-restaurant-vs

Please rate this

Will battery swapping replace fast charging for electric vehicles?

9

October

2022

4.5/5 (2)

While popular carmakers producing electric vehicles (EV) such as Nissan, Tesla or BMW are focusing on how customers can charge their cars in a faster and more seamless way, new entrants in the market are pushing another method to give EV their charge. Founded in 2014 in Shanghai, China, NIO is a car manufacturer specialising in EV (Wikipedia, 2022). Nio is known for developing an alternative to the charging stations: the swapping stations. In December 2021, NIO had more than 700 swapping stations throughout China, which provided more than 5,3 million battery swaps in total (NIO, 2021). The principle of these stations is simple: when an NIO owner arrives at the station, the car parks itself on the swapping location, the battery is swapped within 3 minutes and the driver does not have to come out of the car during the whole process. The business model behind this technology was named by NIO as “Battery as a service”. The car can be bought without a battery, leading to a cost reduction of around 10.000 USD. The customer can then pay per swap or take a subscription that will include a certain number of swaps in a month.

But is this technology better than the existing fast-charging possibilities?

Battery swapping comes with advantages, but also a set of drawbacks.
First of all, the main advantage of this technology is the waiting time for the user. While a full charge for a Tesla at a Tesla Supercharger takes around 25 to 30 minutes (Cline, 2022), a battery swap usually takes between 3 and 5 minutes. A report by McKinsey showed that the waiting time for the charging process was one of the barriers to the adoption of EVs (Heineke et al., 2020). Additionally, this short time required for swapping a battery allows for more efficiency, which would be ideal in congested areas. The second advantage is economic. In swap stations, batteries can be charged when the cost of electricity is lower, making it more cost-efficient. Thirdly, this service can enable easy battery upgrades when a more advanced battery is available. This can help preserve the car’s performance, but also resale value.
On the other hand, battery swapping brings the problem of battery standardization. Not all batteries are the same size, and not all carmakers are willing to standardize their battery size among their models. Moreover, always having charged batteries at the swapping stations require the manufacturers to produce far more batteries than cars, increasing the pressure on the finite supply of resources required for their production.

To conclude, it is still unclear whether battery swapping can achieve better economical, ecological, and practical performance than fast charging. Both methods have great opportunities, but they both require important development. As the BaaS model is increasingly important in China, it will be interesting to follow its performance closely, to understand where and how the swapping stations can play a role in the new EV charging network. It is easy to imagine a coexistence of these two systems, where battery swapping could benefit professional usage, and where charging stations would be used for the regular commute.

References:

Cline, A. (2022, July 22). How Long Does It Take to Charge a Tesla at a Charging Station? MotorBiscuit. Retrieved October 9, 2022, from https://www.motorbiscuit.com/how-long-does-take-charge-a-tesla-charging-station/

Heineke, K., Holland-Letz, D., Kässer, M., Kloss, B., & Müller, T. (2020). ACES 2019 survey: Can established auto manufacturers meet customer expectations for ACES?

NIO. (2021, December 10). NIO Achieves Annual Target of 700 Battery Swap Stations Ahead of Schedule | NIO. Retrieved October 9, 2022, from https://www.nio.com/news/nio-achieves-annual-target-700-battery-swap-stations-ahead-schedule?noredirect=

Wikipedia. (2022, October 9). NIO (car company). Retrieved October 9, 2022, from https://en.wikipedia.org/wiki/NIO_(car_company)

Please rate this

A disruptor in distress – Why Netflix has to innovate to stay ahead

27

September

2022

No ratings yet.

Netflix has enjoyed astonishing growth over the past years, challenging traditional television companies since the launch of its streaming platform in 2007. The company is considered a textbook example of a digital disruptor, confronting traditional industries by employing technology and innovative business models. However, being the first to create a new industry does not guarantee long-term success. In the first two quarters of 2022, Netflix’s subscriber count decreased for the first time in the company’s history. Even though this decrease was anticipated by analysts, the announcement sent shock waves through the stock market, causing Netflix shares to fall 60% compared to the beginning of 2022.

Netflix’s cumulative subscriber count

The loss in subscribers is hardly the cause for this reaction, but rather the underlying problems that were revealed during the earnings call. Facing increased competition from Disney, HBO, Hulu, and the likes, Netflix has to invest heavily in content creation and customer acquisition. In 2018, Netflix recorded a negative free cash flow of $3 billion, followed by a negative free cash flow of $3.3 billion in 2019. Over the same time, the subscriber count increased by around 57 million. Dividing the total cash spent by the number of subscribers gained, we can find the “cost of acquisition” for each new subscriber is around $110. 

So, why do these numbers matter? The simplicity of the subscriber-based business model that enabled Netflix to scale so quickly now poses a serious threat to its ability to grow further. Acquiring new customers by producing and adding new content, running marketing campaigns, etc., effectively costs Netflix more than they gain in extra revenue from these new subscribers. Raising subscription fees will likely deter customers, especially since there are plenty of competing streaming services to choose from. It seems the only way to solve this conundrum is to create new revenue streams by adjusting the business model. 

In November of 2021, Netflix announced the launch of a mobile gaming platform integrated in the Netflix app, featuring a small number of games centred around popular titles such as “Stranger Things”. However, these games are currently included in the subscription fee and feature no ads or in-game payments. In the future, Netflix could lift these restrictions and create additional revenue streams, building on their strong portfolio of original movies and TV shows. Another strategy the streaming platform currently explores is the addition of a cheaper, ad-supported subscription tier that appeals to more price sensitive consumers. 

Only time can tell whether these attempts will be effective in helping Netflix overcome its growth issues, but it goes to show that even industry disruptors have to keep innovating to stay ahead of the competition. 

Sources:

https://www.forbes.com/sites/jimcollins/2020/01/22/netflixs-business-model-does-not-work/?sh=7dfc4d3c22cc

https://www.fool.com/investing/2022/04/24/the-netflix-growth-problem/

https://www.bbc.com/news/technology-59136945

Please rate this

Would you like to know whether it will rain in two hours’ time?

30

September

2021

5/5 (4) Recent research shows artificial intelligence can help us to predict the weather. The system is called the nowcasting system developed by Scientists at Google-owned London AI lab DeepMind and the University of Exeter partnered with the Met Office.

In a traditional way, we use complex equations to forecast for only between six hours and two weeks’ time. The main global issue – climate change makes it difficult to make more accurate and reliable predictions for the weather. Sudden changes in the weather are caused by climate change. The decrease in the temperature difference between the North Pole and the equator challenges our current methods to make accurate weather forecasts. This issue urged the meteorological institutes to develop their tools and methods to follow the changes.

So, there is a big problem that the traditional methods and their tools cannot provide the same performance for the society and community. If we cannot forecast critical storms and floods, it can have more dangerous catastrophic consequences. The recent news all around the world showed unexpected critical storms and floods can cause life losses. This can lead to bigger damage for the society, and to increase the financial costs.

Good news is that! The scientists claim that AI systems can be the solution for this problem! It can make more accurate short-term predictions, including for critical storms and floods.

The research, published in the journal Nature, claimed that meteorologists significantly preferred the AI approach to complete their methods. AI approaches can make the method more powerful by decreasing time for forecasters with predicting the continuous growing data. Instead they can spend more time focusing on gaining a better understanding of the implications for the forecasts. For this reason, the AI approach can help to mitigate the negative effects of climate change and prove better predictions for the people. So, it can potentially save lives!

Sources: https://www.bbc.com/news/technology-58748934
https://www.sciencedaily.com/releases/2019/03/190322105718.htm

Please rate this