Rollo62

Hi there, I'm working cross-platform under Macos Monterey, Win10, VmWare Fusion for quite a while now, without using TPM in my VM's. Now I am more and more forced to use TPM as well and after the current Macos Ventura I face a lot of performace issues, not only under Mac itself, so probably I consider real hardware issues too. Thats why I try to move my projects now from VmWare to Parallels now, which should have a better hardware support and performance for Macos and M1 chips in general. Currently I'm checking out Parallels Desktop Professional and the possible impacts when using it. Generally speaking Parallels and VmWare offer many similar properties and its not too difficult to get it running. Unfortunatly I cannot import my current VmWare VM's there, this always refuses, even Parallels support couldn'T help much. Thats why I have to set up complete clean, new VM's now and I want to ensure they were future-proof (Win11), while my old VM's were not So that means using the virtual TPM-Chip is a must, to allow an upgrade to Win11 later one day. With TPM you cannot easily move or clone VM's anymore, so my workflow needs to be a bit updated then. 'No matter if VmWare or Parallels, when you move to another host PC, you have to transfer not only the VM but also the TPM-Password. Here is a nice explanation how a transfer from one physical Mac to another Mac can be done, but the question is not Mac-related only. Maybe someone had faced these TPM topics before and found a clever way howto ensure that VM's can be cloned and transferred without big hazzle, even if it includes TPM ? Best of course to switch on/off the TPM, wherever needed. Of course a "moved" VM has same MAC-Adress and UUID, so there should be only one VM active at the same time in a network. Interesting for me is to clone with separate "MAC-Adress and UUID" to testing, which could be used for fast, separate test, but it should not crash the whole network than if accidentily running both. Yes, I should use Snapshots for that, but I want to avoid to clutter and risk damage to my original VM, thats why I prefer complete cones instead. Another use-case is to make clones from a general template VM, all this won't be easily possible with VM any longer. How do you organize your VM-ecosystem including TPM, is there still a way to workaround TPM in VM's ?

VmWare VM is not working under Parallels for me, at least not the ones I want to. There is a conversion tool, right in the beginning, which immediately fails with little or no errors. Even the Parallels support couldn't solve that, so I have to re-create a clean Parallels VM anyway. More headache to come, it seems that Parallels VM on Intel cannot be ported easily between Parallels M1. Only exchange by copy disk data is a way they provide. https://kb.parallels.com/125344 So then a breakdown of Macos-Intel will require a brand new VM-Setup under M1 anyway, and M1 will be the next Macos for sure, as I think Macos-Intel will be deprecated. Parallels VMs are not interchangeable between Intel <-> M1, no matter if TPM. Copying of harddrives should be easily possible, if not using Bitlocker ( encrypting by TPM ), which I do not consider to use in my VMs, so thats OK for me, but I thought the TPM will encrypt or certify some parts of the Windows-OS. https://www.linkedin.com/pulse/windows-11-requires-tpm-what-why-matters-rand-morimoto Ok, at least the VM data seems to be safe by direkct access of the hard drive files, but the OS and probably all installations and configurations of applications seems to be lost. Microsoft says https://learn.microsoft.com/en-us/windows/security/information-protection/bitlocker/bitlocker-countermeasure Is this meant as an option, when Bitlocker is activated ? From my assumption TPM is doing the latter, no matter if Bitlocker is activated, or not. If TPM recognizes any changes, then it prohibits booting the OS and I assume this cannot be reset somehow. It seems my TPM-less days will be gone soon and I have to optimize the process of setting up a new, clean VM ( which I already did ). At least the cloning and copying from "template VM" seems to work still on the same machine. From my tests I can use the Parallels "Clone" function, as well as the copy-and-paste of the *.pvm images, to create a derivate of my template VM. The derivate VMs seem to stay activated, as far as I can see. Conditions: No Bitlocker used, No more than one of the "cloned VM" active at a time

With some experiments, it seems that it doesn't work so seamlessly as described in the Parallels description: There is not only one entry storing the TPM password, but there were two ( /System and /iCloud ). Both contain the same Password, when I extract that from KeyChain, looks similar to this: 0a9dfe9dbabcdfgdsddf4678fc3defsf5fs5fs67hd7n8899f086b795cc9a1509c3 The only difference seems what is so called "Location" - System: One time with Account number added - iCloud: One time without All the rest, including password seems identical. When I try to copy the entries into my own, custom KeyChain Bundle, then I can only copy one f the two versions, so I assume "System" is the right type. IMHO the best way to copy that, is - Unlock the "System" and "Your Custom" Bundle by right click - Unlock - Select the left "System" KeyChain Bundle - Enter the "TPM" in the search field - Sort by the Key type "System" - Select and right click copy all the "System" entries - Ignore the "iCloud" entries, they seem to be not necessary - Select the left "Custom" KeyChain Bundle - To right click and insert them all into the custom bundle - The Custom Keychain Bundle "MB02_Pls_TPM.keychain-db" is for example stored under /User/Library/Keychains, and can be copied to another machine. - Such KeyStore could be Added to another machine hopefully - From there, on the new macnine, the TPM could be copied from "Custom" to "System" to make the VM work again Is there a better way to it ?

Thanks for the info, I have two Macs, wile only one is for current development. I also want to ensure that I won't end up in desaster when this maching breaks down one day, if I would consider TPM too late this will break my neck. A breakdown of one host machine might breakdown 10 guest-VM too, I really get scared about that scenario and I want to do the right thing before its too late. Regarding the workflow, I hope that still cloning variant, particular VM's from a generel "template" VM will still be easily possible. Usually I use only one particular VM at a time, for lets say VM-Delphi, VM-WebDev, VM-VisualStudio, VM-TestEnv, ... so clones from one template VM was no problem in the past. On one host machine I think that should be OK, since all the VM's relate to the same certificates set on that host machine ( at least I hope so ). If I have a few VM's related to one host in that way, what happens if the host breaks down completely, How do I prepare to be able to move quick and fast to a new machine then ? I think backup of a VM must always contain the VM-image plus its VM-TPM-Certificate then, to be able to restore it. On the Mac it seems that its possible to copy the KeyChain-Bundle as collection, like in my link above, which could be exchanged between different Mac's. Maybe I should bundle all the TPM certificates to that single bundle in advance and backup this bundle. That looks like a possible, clean way under Macos-Parallels, I'm not sure how it might look under Macos-VmWare or under Windows-VmWare hosts. From yur example under Windows it is called "Shielded VM Certificates", which is maybe the way to use this under Windows. Anyway, I see no way to exchange VM's interoperably between different hosts structures in the future, like moving from Macos-VmWare to Windows-VmWare or into the VmWare-Cloud and vice-versa. To be honest, I never really used that feature aside testing it, but it was always good to have it.

Me neither But I think it should be fairly simple to get the Python4Delphi bridge running and to understand the basics, while I assume that on the python side you may choose from a bunch of different Voronoi library incarnations for testing. To call such voronoi functions should be similar in each case, if you have the boilerplate code running.

Have you considered Python4Delphi, I would assume that Python has all the modern libraries easily available.

You can try to create a new, empty app in XCode, with same bundleId net.mustangpeak.LccThrottleApp, and try to set up the provisioning profiles for debug and release from there. Usually XCode is able to set up the provisioning right and at least can give some more useful hints if not. Sometimes older or several, duplicated certificates were in the KeyChain Access, which need a clean up before installing the new ones. If XCode can build and run an app then usually its OK for Delphi too.

Interesting question. I would think that the problem is that your app sleeps in doze mode, and there is no real reason to permit it running. When in foreground then the TForm can read those key events, but in background you could try to use an Android Service. Maybe this method works still within a service, but I doubt that you will get an background permission easily without certain hacks. Maybe you can look after howto receive the volume key from music player apps in the background, there were some possibilities too, but the main difference seems to be that a music player app is allowed to run in background. You must consider that reading keys in background would be a large security issue (like rading passwords from external keyboards), so I would think Android strictly forbids that.

Nope. I think that chatGPT is trained and powered from SO, beside many other sources. So it should not be able to replace its original sources, since the global "brain" of millions of human developer brains is simply to valuable for new input and failure corrections.

It should be able to uninstall with something like this Prints all packages adb shell pm list packages also show the package UID adb shell pm list packages -U also include uninstalled packages shell pm list packages -u ... remove this app package from the device adb uninstall test.apk Keep data and cache directories after removal adb uninstall -k test.apk

Are you able to connect via ADB and you cannot list or delete the desired package name via ADB ? Maybe reboot of the phone could help, as sometimes phones can stuck somewhere while they look and feel normally operating.

Do you want a bloody copyist ?

Don't miss these books ...

Yes that's true, so what ? Yes, I'm not an AI engineer working at OpenAI, so I am probably a complete AI noob. But I know enough about AI, neural nets, neural processors and backpropagation to make my thoughts about it. At least I'm not alone in thinking a kind of "AI consciousness" could possibly arise from a critical mass of "neurons" and data, these theories exists I think already since the 60's also from renowned scientists. I only said that chatGPT and other AI's point clearly into that direction that this could be the case, from my point-of-view. I also might add that of course an AI will never be able to mimick a human brain, since it works on completely different hardware and structures, but both brain and AI were investigated and knowledge combined heavily in the past years, there have been too many breaktroughts to count them here. Would it make any difference if AI is perfectly "mimicking consciousness", or not, if it probably has the same outcome as from a human brain's "consciousness" ? Moreover I pointed more to improving the usefulness of AI than to enforce "AI consciousness", which is not much relevant to me if the output is good. The "AI consciousness" lays in the astonishing creativity of writing lyrics, painting images, making music, all this works astonishing well, you should be able to agree to that fact. Yes its only a clever tool, sometimes we see complete rubbish, but sometimes we can harvest some pearls, what so wrong with it ? If you can get the same output with some Delphi classes, then I completely follow your words. I'm off now, I see so much negative energy against, instead of looking at the current AI possibilities and possible future optimizations. Future will tell.

Last one in uses wins.

Sure, I cannot read anything about chatGPT, although I think its based on DaVinci model as well. Even the usual GPT-3, DaVinci Playground points clearly to the particular chatGPT playground. Here is the reference to that link as well https://openai.com/blog/chatgpt/ Not the usual playground. I also played with the GPT-3 API, same as you do, it seems that chatGPT was especially trained, not the usual GPT-3. Lets wait and see whats coming up ...

I'm sure that if you focus on a special problem only and tailoring the questions and prompts very well, you will get reasonable answers. Ok, not 100% reasonable, maybe 95%, but I have seen very many examples where it works out really well. This depends on the topics and domains very likely, not in every domain this might work. Anyway, you have to check out for yourself. I remember well when I've got my first Sinclair Z80, programming one of the first programs, I've programmed a simple Eliza program. This was very simple basic program of coarse, but from that I realized how easy text processors can lead to interesting Q&A. And yes, Eliza was completely useless and giving not really helpful answers. The AI developments and technology were sleeping ever since on a low level and only in the last few years some significant, exponential changes happened in that field. I know the power of exponential growth and that it might show reasonable results soon. So many people nowadays are getting hot on AI and working on that topic, don't you agree we will see a real chat or human-like AI soon, by fixing the last 5%-10% of issues ? I do and I don't say that chatGPT is the last step of the evolution, its only the start.

Maybe its like that, but I doubt that this is 1:1 the same engine, because GPT3 is already there for years and chatGPT has a much evolved quality. Yes, I think its based on the same engine maybe only differs in the training data, who knows. Nevertheless, I'm afraid we have to wait until OpenAI reveals the secrets and the correct API. Or do you get the same answers from your API and the Playground ?

@Anders Melander Maybe its a question of different expectations ? 1. You might expect answers like from a nobel price winning professor, which is fair enough, so that you can use the data 1:1 as is. 2. I would expect answers like from an half-educated sales assistant or a passenger on the street, giving only 25-50% of useful information but leading into the right direction, which leads to useful thoughts from where one can search deeper if he likes. For example, from above examples, I was not aware that Wilhelmshaven has a sandy beach nor that it has a Jade Stadion, not that Giza was build from Granite and Limestone. Clearly 1.) is the most desired goal, but 2.) is good enough as a tip-starter for so many cases. That's why I would'nt say that 2.) is completely useless, I see enough use cases for both options in many fields. I would predict that the answers from 2.) running through a better "data validation" process in the near future will solve most of the issues complained about.

Just for fun, I've tried to pull some "unsolvable" information out of chatGPT, where I think it has given well suited answers to my questions. ( It remembers that we agreed on the name Assistant a some days before ). Ok, the math and results were a little awkward, but tending towards a reasonable direction. If you consider that this is only a text processor, then this outcome is amazing (at least to me). To implement some formula and data verifier should be maybe not that hard, to verify hard facts like limestone density, etc. I think the text processor is best if its not trying to answer hard facts and math right now, but this has clearly been an huge evolution to GPT3 1-2 years ago.

I can only say that chatGPT can give really helpful answers and amazing "understanding" of facts, no matter if it really understands or not. Of course it can have its errors still, but it most of the time completely follows and understands your intentions, even if you ask not so well formed questions or try to trick it somehow by slang language. When you really ask well defined questions about very specific topics, like for example for craftsmen tasks at worksites, how to best solve their tasks in a given situation, it can give very good advices in 98% of the cases I would say. This is what I would also expect from a human assistant or teacher and to be honest not many of the shop assistants can help customers questions that adequate. I would rate chatGPT about 8 and the average, unmotivated human shop assistant 4-6, on a 0-10 scale. Please try chatGPT really out in practice, before you throw it into the "irrelevant" bin so fastly. I'm very sure that we all have to deal with this kind of AI in the near future, a few years from now. At least I have a lot of ideas howto use such AI things in the real world.

Are you sure you talk to chatGPT, or is it GPT-3, its predecessor which is for quite some time out there ? https://en.delphipraxis.net/topic/8084-chatgpt-example/?tab=comments#comment-68216 Unfortunately OpenAI keeps the API a little unclear, you can check the results with the chatGPT Playground to be sure. The real link to chatGPT playground is here https://chat.openai.com/chat

Thats a fair analysis and I completely agree. But if you see whats possible with those AI, like the LETA series from Dr. Alan Thompson, is really amazing. He has done a lot of work with various AI's, and knows how to optimize the parameters and prompts. And yes, its still only a text processor, like an advanced RegEx processor maybe ... 🙂 Since chatGPT is even more advanced than LETA, as it may remember older prompts, I expect even much better results. You will always have your point that its only a text processor and not a human, but the chats you can do with those machines are quite inspiring too. According to the Turing test, a computer can be said to possess artificial intelligence if it can mimic human responses under specific conditions. If you cannot really proof if its human or an AI in a normal chat, wouldn't it be fair to call the AI somewhat "intelligent" ? What would be your proposal for a proof of creative "intelligence" ?

Oh yes, then better upgrade to D11, its worth it.

I assume that he has the D11 code, so it should be legal to rework this unit to make his code-base backwards-compatible. If it would be true what you say, this would forbid to fix any buggy system unit for the same reason.