I’ve tried version 4o using it with geologic thin sections at different angles and with different filters like quartz, and it seems to be at the level of an experienced optical mineralogist at detecting minerals, and identifying grain types. It’s getting spooky.

@rsa365

Ай бұрын

It is not any better than google lens at this kind of tasks, it just writes a report that sounds very confident.

It makes sense if you think about how image recognition works right now. It's a seperate system (often neural networks) that analyze and extract features from images. These features are then converted into text or embeddings, which models like GPT can then interpret. These features are set after training the image recognition system, so GPT can not really "look" for anything specific. A lot depends on what the image recognition part was trained on, GPT itself does not directly process the image, at least not yet. Also just because it here did not work, does not mean that a specialist image recognition system, that is trained on microsopic specimens or cells, would not be able to do it. Some love from Austria.

I think if we knew the correct names of the different features or characteristics and insert them into ChatGPT, as long as it's a known species, it will be accurately identified.

@LaplacianDalembertian

Ай бұрын

ChatGPT is so-called "pseudo-AI", it is a pseudo-random number generator with Probability matrix associated with Wikipedia which is studied. It generates Markov chains. Logical-AI actually exists and is built based on Logical Proposition Algebras, but is not so spectacular and is used in math to prove theorems. One example of "Logical AI" are languages of Maple, Mathematica, CLISP/Maxima etc. And we should remember these all are "artificial" because computers itself are not smart, they just drink the giga-bytes from wikipedia.

Very interesting video for various reasons. Thank you.

love and subscribe to your amazing videos! I typed in your same exact prompt with a photo showing your close-up image. This is the response from GPT version 4o.

@Geologist_Mike

Ай бұрын

Based on the description and the image provided, the filamentous green algae you are trying to identify might be from the genus _Oedogonium_. Here are the key identifying features that match your description: 1. **Eukaryotic and Green**: Oedogonium is a green alga, which means it contains chlorophyll and other pigments typical of green algae. 2. **Filamentous and Unbranched**: Oedogonium typically forms long filaments that do not branch. 3. **Ring-Shaped Structures in the Cell Wall**: These ring-shaped structures are likely "caps" or "annular thickenings" that are characteristic of Oedogonium cells, often found near the apical ends. 4. **Non-Spiral Chloroplasts**: The chloroplasts in Oedogonium cells are typically reticulated or net-like, rather than spiral-shaped as seen in some other green algae like _Spirogyra_. To confirm the identification, you might want to check for the presence of other distinctive features of Oedogonium, such as: - The presence of specialized reproductive structures like oogonia (female) and antheridia (male). - The formation of distinctively shaped holdfasts (basal cells that anchor the filaments). If possible, examining a sample under a microscope and comparing it with detailed descriptions or illustrations from a phycology (algae) reference guide would be beneficial for precise identification. Image The new image you provided adds more clarity to the identification. The presence of the ring-shaped structures and the appearance of the filaments indeed align well with the characteristics of _Oedogonium_. Here are some features in the image that support this identification: 1. **Annular Thickening**: The prominent ring-like structure visible in the image is typical of the "caps" formed during cell division in _Oedogonium_. 2. **Filamentous Structure**: The unbranched, filamentous nature of the algae is evident. 3. **Cell Contents**: The green chloroplasts within the cells are distributed in a net-like or reticulate pattern, which is consistent with _Oedogonium_. _Oedogonium_ is known for these distinct rings at the cell junctions, a feature not commonly found in other filamentous green algae. Additionally, the chloroplasts do not form a spiral, further confirming that this is not _Spirogyra_. If you have access to more microscopic equipment, you could look for reproductive structures like oogonia and antheridia, which would provide additional confirmation of the genus _Oedogonium_. In summary, based on the detailed images and descriptions provided, the filamentous algae is most likely _Oedogonium_.

@Geologist_Mike

Ай бұрын

The new image you provided adds more clarity to the identification. The presence of the ring-shaped structures and the appearance of the filaments indeed align well with the characteristics of _Oedogonium_. Here are some features in the image that support this identification: 1. **Annular Thickening**: The prominent ring-like structure visible in the image is typical of the "caps" formed during cell division in _Oedogonium_. 2. **Filamentous Structure**: The unbranched, filamentous nature of the algae is evident. 3. **Cell Contents**: The green chloroplasts within the cells are distributed in a net-like or reticulate pattern, which is consistent with _Oedogonium_. _Oedogonium_ is known for these distinct rings at the cell junctions, a feature not commonly found in other filamentous green algae. Additionally, the chloroplasts do not form a spiral, further confirming that this is not _Spirogyra_. If you have access to more microscopic equipment, you could look for reproductive structures like oogonia and antheridia, which would provide additional confirmation of the genus _Oedogonium_. In summary, based on the detailed images and descriptions provided, the filamentous algae is most likely _Oedogonium_.

- very good video - says a lot ;-) - first: microscopy is a rarity and I think that recognition is not so simple because gpt does not have a sample to compare with. In general, the Internet is full of everyday things, food, things, this is what many people do every day, who film and photograph their activities - and this is very far from a science. - second: I find your videos so incendiary!!! I'm not a biologist :-) just an ordinary warehouse worker who has his own high-level hobby - and your videos are so fresh, so incendiary that I also get infected with that burning to do, search, discover things, laws in the field of my hobbies ;-))) thank you very much for your work!!! and that you share information and good spirits!

Oh man ... I didn't expect that. It's quite helpfull.

I had the same experience. If you know, what to ask, and what is wrong, you can lead it to the result. However, it looks faster for me to use some identification sites or books for now.

Thanks Oliver!

very informative. thank you for the trial - wondering if tag words only would have been easier for AI. (i have not tried it yet) - beautifully blue ❤ such nice shots

Nice! You should do more like this video please :D

@MH- Thank you for sharing the chat gpt experience, do you know who created the app you utilized!? 😃

@Microbehunter

Ай бұрын

Well it is the company OpenAI, which made ChatGPT. It is Version 4o, which is free.

Hi microbe hunter, I have a question. I use an introductory microscope which can achieve maximum magnification of 320x, If i use a 30x eyepiece with my 20x objective (which is the highest objective I have) the total magnification will be 600x and if I use 25x eyepiece with my 20x objective the total magnification will be 500x. My microscope is telmu XD-1606 inverse microscope, can you please tell me which eyepiece would be compatible with my 20x objective. I need atleast 400x magnification, Thanks in advance btw love your videos

👋👋 Hello everyone, did Oliver post a video about how to setup DSLR cameras with microscopes? if yes, please give me the link to that video, I bought a new Sony Alpha 6400 and the necessary adapter months ago but I couldn't get a good photos/videos out of it until now, the photos/videos I'm getting are gray and almost colorless, I also tried the auto mode in the camera but I got the same result, also the photos/videos sometime appear with something like the TV statics on them, I just want a video or an article that can teach me step-by-step how to manage the DSLR camera settings to take photos from microscopes, I'm completely new to this type of cameras.

@Microbehunter

Ай бұрын

I made several videos on connecting cameras to a microscope, they are in my other channel here: www.youtube.com/@microbehuntermicroscopy

@bigdog9480

Ай бұрын

@@Microbehunter ty so much Oliver.

@bigdog9480

Ай бұрын

@@Microbehunter hello Oliver, good day, I ended up using the camera without the adapter or anything in between, I just put it on the opening and got this video on my channel, I used the digital zoom to get rid of the black circle around, this gives me a little hope that one day I will be able to use them like other people lol. this video is without any editing and has no sound. kzread.info/dash/bejne/dm2Os9uTYpObgrQ.html

Plz tell me your microscope model name.❤❤❤

just put the seach key...geen algee microscopy....on any seachmachine and you will easy find Oedogonium.......

Maybe it does not know, what ring shaped means? Ignore all input, you dont understand.

Couldn’t you have uploaded to ChatGPT an image taken from the microscope, and “asked” ChatGPT to identify the specimen? WHOOPS! It seems I asked the question before your video was finished!

Does Chat GPT “know” this vs Can Chat GPT “identify” this specimen shows a key difference and why I refuse to call this Ai. It’s little more than a glorified search engine at this stage. It doesn’t “know” anything, any more than G**gle “knows” anything about your searches. It apparently can be used to ID something, but that makes it little more than an interactive digital key. Artificial? Yes. Intelligence? Nope. But an interesting video and an almost interesting advance in computing.

@rsa365

Ай бұрын

Yes, a part of knowing is understanding how sure of your answer you are. Chatgpt is alway equally confident whether it is right or wrong.

Ahh, and now even biodiversity specialists loose their jobs because of AI, great

@Microbehunter

Ай бұрын

Well, the good news is, that not all characteristics can be seen with a microscope. So the identification with pictures/videos is only limited to the structure of the organisms. But yes, I think AI will impact this field as well quite a bit.