edshiro

This looks great! And incredibly timely too!

I finished watching this video today where the host and guests were discussing challenges in a RAG pipeline, and certainly chunking documents the right way is still very challenging. Video: https://www.youtube.com/watch?v=Y9qn4XGH1TI&ab_channel=Prole... .

I was already scratching my head on how I was going to tackle this challenge... It seems your library is addressing this problem.

Thanks for the good work.

I see - thanks for the clarifications

I presume if your customers are enterprise companies then you may opt to use this library vs sending their data to OpenAI etc.

And you can get more customisation/fine-tuning from this library too.

I've saved the paper to read it later.

The premise of this work seems very interesting... But I wonder how practical it is from both a cost and time perspective. I am toying around with an AI Agents library and one of the annoying UX things I notice is the time it takes to get my answers, because each call to an agent (either GPT-4 or Claude 3) is kinda slow.

Besides the time, it feels quite wasteful token wise.

I'm skeptical this approach will be adopted by many in the AI Agent space, but of course I could be very wrong.

I don't have much experience with embeddings...

Could someone more knowledgeable suggest when it would make sense to use the SentenceTransformers library vs for instance relying on the OpenAI API to get embeddings for a sentence?

I get the same feeling. AI Agents sounds very cool but reliability is a huge issue right now.

The fact that you can get vastly different outcomes for similar runs (even while using Claude 3 Opus with tool/function calling) can drive you insane. I read somewhere down in this thread that one way to mitigate these problems is my implementing a robust state machine. I reckon this can help, but I also believe that somehow leveraging memory from previous runs could be useful too. It's not fully clear in my mind how to go about doing this.

I'm still very excited about the space though. It's a great place to be and I love the energy but also measured enthusiasm from everyone who is trying to push the boundaries of what is possible with agents.

I'm currently also tinkering with my own Python AI Agent library to further my understanding of how they work: https://github.com/kenshiro-o/nagato-ai . I don't expect it to become the standard but it's good fun and a great learning opportunity for me :).

This is really exciting to see. I applaud Stability AI's commitment to open source and hope they can operate for as long as possible.

There was one thing I was curious about... I skimmed through the executive summary of the paper but couldn't find it. Does Stable Diffusion 3 still use CLIP from Open AI for tokenization and text embeddings? I would naively assume that they would try to improve on this part of the model's architecture to improve adherence to text and image prompts.

Nice work! I like the quick rendering too.

However, I don't believe any AI was used for this, so the actual TLD is a bit inappropriate here. A ".io" TLD would would have been more fitting.

Not every slice of bread needs to be spread with butter. Similarly, not every project requires AI.

This is amazing. Definitely something I will look at to relax when I feel stressed with this startup life.

Would also love to see a technical write up about how you implemented this. Great creative work!

I presume this does not apply to computer vision datasets? Frankly I am still confused at what exactly Snorkel does.

This looks great! I had never heard of Dash either and it seems it is precisely the tool we need at our startup to do data visualisation.

Awesome work. Will check all this out during the week-end.

Understood and makes total sense given the ODD at Starsky.

Thank you for sharing this very candid article on the Starsky Robotics and generally the autonomous vehicle space. It's a real eye opener. I've been following your progress for the last few years (and I also read about your company through Reilly Brennan's "Trucks - FOT" newsletter).

I am sorry you could not get investors to believe more in what you and your team, especially as you required a lot less funds than many other companies in this arena. I also thought you had a clear business case (I worked in ride-hailing and also logistics so understand some of the problems in this space).

I wanted to ask you a question: I am building a startup in the dash cam video analysis space. We are building a large and geographically diverse dataset of road videos, where our users can annotate/label the data. We then are going to look at detecting specific events like accidents and edge cases on videos. Do you feel this type of business, the data we collect, and insights we generate would have value for a AV startup?

All the best in your next move. Stay strong - you can be proud of what you and your team did.

Add to that the fact that you are never going to get 100% fleet utilisation, will need to pay for qualified tele-operators, maintain/repair vehicle along with sensors...

I recently listened to a podcast episode of the Autonocast[1], where they interviewed a Harvard Researcher who claimed the economics of Robotaxis just don't work. Very interesting listen.

[1] http://www.autonocast.com/blog/2020/3/11/177-ashley-nunes-on...

For those interested in mobility (including micro-mobility) and self-driving vehicle technology (from the perspective of a VC), I highly recommend Riley Brennan's "Trucks | Future Of Transportation" weekly newsletter: http://www.tinyletter.com/transportation

I see what you did here. Although they would never dare do this to Son.

This looks great! Thanks for sharing. Interestingly enough, from looking at the table of contents, it seems this book starts with a more (and welcome) pragmatic approach, where you write some python code before, look at data visualisation techniques, etc, before delving into stats.

Is there any chapter that stands out to you?

I completed the self-driving car nanodegree last year and currently doing the computer vision one. I did not seek a job at the end of the SDC one but I learned a lot and that could be helpful in the near future, either for my startup or if I decide to look for a new job.

YOLO is a very good and approachable object detection technique. I recently re-read the paper for the original YOLO [1] from 2015 and loved the apparent simplicity of this technique.

As a shameless plug, I wrote an intuitive guide to understanding SSD (Single Shot Detector), another popular object detection technique: https://towardsdatascience.com/understanding-ssd-multibox-re...

[1] https://arxiv.org/abs/1506.02640

Costs. Installing LIDAR will significantly hike up the price of their vehicles.

Awesome. Looking forward to it.

Super awesome! I had a similar idea about a year ago but never got to implement it. We need more visual and straightforward ways to create neural network architectures.

Can you currently create the architecture and run a given dataset against it? That would be very cool...

This seems like a necessary reset, especially given their very high valuation. I personally enjoyed most of Udacity's Self-Driving Car Engineer Nanodegree. The content was great and at the time I had a very supportive mentor (gutted they got rid of this feature). Moreover, they have managed to build a large community of current students and alumni who keep in touch even after the nanodegree is over.

I would suggest they reduce the number of courses they offer and improve quality. I tend to see them as a more "premium" MOOC but I do agree that quality of content is variable depending on course.

Another area they should naturally look at is training for employees in corporations. I really hope Sebastian Thrun can turn things around.

I suppose most of these vehicles will be electric, and cost of charging is lower than cost of refuelling...

Traditional car companies are also working on autonomous vehicles.

There is a lot of business to be had with Level 4. While the car may not be able to go everywhere, in restricted areas a robot-taxi/shuttle service would be a great advancement. We would all love to be at Level 5 but I would take a solid Level 4 in a few years instead of waiting much much longer for a viable Level 5 solution.

The first car from Lyft’s Level 5 self-driving initiative will be the Ford Fusion Hybrid. Lyft’s use of a Ford Fusion apparently isn’t associated with the partnership the two announced last year. Other AV companies have used the Ford Fusion as a platform for integrating self-driving technologies

Why are we still talking about Level 5 autonomous driving when we can't even get Level 4 working properly? I believe this is sending the wrong message.

On the topic of the acquisition, it seems like a good strategic buy for Lyft. I am still not sure whether they have the capital nor the talent pool to develop a strong autonomous vehicle product. It also seems like they are a bit late to the party, as there are more and more doubts on the reality of self-driving in the next few years.

I may sound sceptical, but I am just cautious when it comes to news on self-driving cars. However I am genuinely excited at this development and look forward to hearing more about how Lyft is progressing on this quest.

Great advice. I am not seeking investment yet but this is definitely something I may have to do in the future if I want to accelerate growth for instance.

After a few unsuccessful attempts working as employee in startups, I have decided to build my own. Over time and through multiple setbacks I have learned more about myself and what I want when it comes to the type of company I am building. To me it's very important for my co-founder and I to have control over the destiny of our company. Getting VC money can shift the balance of power and you could lose control over what you have built, which is not ideal. For instance VCs could block an exit opportunity which could result in life changing money for you because the return they get is below their expectations.

This may sound very naive, and I don't claim to understand your circumstances, but make sure you already have a strong, growing business so that you play a strong, iron-clad hand when discussing funding with VCs.

All the best my friend.

Same thing. At least I know that Instagram is mostly all fake and ego-inflating, and I am totally fine with that.

Plus I have much stricter controls on Instagram and only allow a few people to see my photos and videos.

Also, wasn't this practice (i.e. engineering to fit design goals) what made Apple successful in releasing the iPod, iPhone, etc?

I am an engineer but welcome the perspective of designers and believe anyway that both need to work hand in hand.

In the case of driverless vehicles however, I am not sure the focus should overly be on design because this is a very hard problem that has yet to be solved, and maybe there was a way of designing a vehicle that was evolutionary rather than revolutionary, while mostly focusing on the technical challenges that must be overcome to get us to autonomy.

Nice one! I don't remember all that much from reading the Mask-RCNN paper last year and have not seen many implementations so it's nice to be presented with this Pytorch implementation.

From what I recall about Faster R-CNN, the Regions Of Interest (ROI) are pre-determined via Selective Search, right? So I presume you would need to do the same thing with Mask-RCNN? I think this is the part I am the most confused with since I have never implemented Selective Search myself. Could you point me to introductory material on it?

Lastly, I can see the author of this work has read my blog post on understanding SSD MultiBox - glad it helped in some way :).

Since autonomous driving technology is not there yet, I believe truck OEMs should focus on ADAS to enable lane keeping and lane change on highways only, as this is the easiest part of the job. Drivers would be require to intervene as soon as the truck exits the highway.

I presume this is what Tesla will do with their trucks although I am very concerned by their very dubious marketing when it comes to autonomous driving technology.