Beep boop - this is a robot. A new show has been posted to TWiT…
What are your thoughts about today’s show? We’d love to hear from you!
Beep boop - this is a robot. A new show has been posted to TWiT…
What are your thoughts about today’s show? We’d love to hear from you!
That pessimist archive is a great find. I wonder if the surgeon generals warning is doing exactly what it was intended to do… Get us talking about the topic? Realizing that teens are consuming this media and interacting with their peers in previously unknown ways
What if the imaginary Microsoft Silicon was a neural coprocessor optimised to work with ChatGPT, capable of connection to a system via USB 3.0? A sort of AI dongle if you like.
It would need to be a Thunderbolt 4 or USB-C 4.0 dongle, at the very least, to get any sort of meaningful throughput.
If you can push enough bandwidth to run external graphics, you could find a way to make it useful
Paul Smith-Keitley
Adobe Creative Educator
The external graphics adapter that run over Thunderbolt 4 has a maximum of 32gbps they can use. Performance of an external high end card is about 1/10th of what it would be, when attached directly to the PCIe bus internally on the PC. (USB prior to 4 doesn’t support Thunderbolt, so would be unsuitable for such an external co-processor, only high-end laptops with Thunderbolt 3/4 or USB C 4 ports would come into question.)
For laptops, where adding an internal graphics chip (or in this case a neural processor) isn’t an option, it is a case of “better than nothing”, but if you are using a high-end card, you are wasting up to 90% of its potential. Thunderbolt uses a total of 4 PCIe lanes, which it shares with everything attached (display, camera, keyboard, mouse, headset and Ethernet, in my current use case), a typical graphics card uses 16 lanes.
It could only be a short-term solution for people who haven’t upgraded their Thunderbolt 3 or USB 4 capable laptops to 14th generation Intel or current Qualcomm or future AMD based laptops or desktops (not many desktops have a Thunderbolt port).
It wouldn’t make much sense, other than as a developer box, to kick start development. But in that case, something like the Microsoft Windows Dev Kit 2023 (an ARM based desktop, similar to the Mac mini) would make more sense.
That assumes we use current ideology, the idea of a local dedicated GPT processor running simple tasks isn’t out of the question, also, not everyone uses laptops
Paul Smith-Keitley
Adobe Creative Educator
But only high-end laptops currently have Thunderbolt, which provides the bandwidth and DMA access an NPU would require. Low-end laptops and a 99% of desktops would be out of luck for such an add-on processor and the performance on those laptops that do have Thunderbolt would be disappointing, compared to a new laptop or PC with a dedicated NPU.
Otherwise, you are talking not about an NPU add-on, but a dedicated Neural Computer that is attached to USB and has its own OS, storage and RAM, which would probably cost as much, if not more than a PC or laptop with a processor with NPU co-processor.
It would be possible to create a. MicroOS, capable of manipulation of a GPT query
I must say that the case of Elon Musk taking Twitter out of the voluntary code of disinformation thing is one of the few situations where I agree with him. Of course, where the information is demonstrably a lie, then this should be taken down, but I don’t want the government to be able to take down anything they don’t like, or that goes against the information they want us to have access to. I would prefer it if all the conspiracy theories and misinformation stayed up and they taught you media studies at school so you could judge for yourself what is true or false; you wouldn’t want the UK, for example, to suppress information about a hypothetical Tieniman Square type massacre under the guise of misinformation.
I look forward to seeing what Microsoft and Intel will do with their NPU stuff. I think we all know that social media harms children, but they often warn us about horror movies, video games, and social media causing violence in our society.
Can we have more from AI Leo, please? Maybe he could host TWiT.
When I get a new iPhone, I don’t restore from the cloud, but I will always set it up as a “new” iPhone, so I’m not sure how many people will use the Windows 11 cloud restore feature.
Yes, you could use a basic Linux installation, for example, although that would be a poor showing for Microsoft, it would have to be some for of Windows Core. But you are still left with the problem that the solution would end up being as expensive or more expensive than buying a new PC with the capability built in.
It depends, it isn’t the government asking for the information to be taken down, it is the law - for example it is illegal to glorify Nazism in Germany, Mein Kampf is banned and the showing of the Nazi symbols in anything other than a historical context (E.g. Wolfenstein had to be released in Germany without the swastikas on the walls), it is illegal do post about hollocaust denial, as is posting about racial hatred or incitement to cause harm or death to others. That is simply a cost of doing business in Europe.
If Twitter finds such posts or users report such posts, they have to be taken down within 24 hours. That is the law.
I reported a post a few weeks back, where some idiots in an SUV forced a cyclist off the road at high speed and he smashed into a tree, all the time, they were laughing & the post encouraged others to go out and do the same thing. The account was banned.
They have the IOT core that runs on RISC cpu
I see your point, Not entirely sure where I was going with that
The USA has a very loose definition of free speech, but it is only guaranteed what the government can and can’t prohibit.
In Europe, especially Germany, social integration is important, as is not forgetting the past - or at least the lessons learnt from it. That is why free speech is still heavily upheld, but within limits that don’t cause harm to others or glorify the darkest chapters of the past.
Germany, especially, went from an evil dictator responsible for millions of deaths, to an opressive regime that spied on its citizens (in the East of the country). They never want either to happen again, which is why the constitution and laws make it pretty much impossible to use CCTV or webcams in public spaces, for example.
Many countries’ laws seek to restrict the freedoms of the population, and whilst some German laws do - motoring law, for example, what speeds you can drive on what roads etc. - many of the laws are about restricting the government and business from interfering in the private lives of the citizens.
Many talk about the over regulation in Germany, that they have laws for everything. They do, but they are often there to actually ensure the citizens maintain their freedom.
In the 80s, a German friend of mine did an exchange year in Los Angeles, his guest family lived on the edge of Beverly Hills. He was shocked at how many freedoms he lost, when he went to the USA - he could no longer drink or smoke (he was 16 & could legally buy alcohol and cigarattes in Germany), especially on school grounds. The Principal allowed him to leave the grounds at break time to smoke, but he was told off for smoking on the opposite side of the road from the synagogue, that was disrespectful, he should and sit on the steps of the Catholic church and smoke there instead (erm, isn’t that just as bad?).
At the guest family’s house, he wasn’t allowed to smoke in the house, so he went for a walk in Beverley Hills and was pulled over by the police, because residents were nervous that a young white man, dressed in white jeans and a white shirt, was walking in the neighbourhood! They told him to go walk in the forest behind the houses instead!?! A couple of days later, a mounted policeman on his horse came along the trail and spotted my friend, “ah, you must be the German exchange student, have a good evening.”
3pm GMT 5/6!23 Bing channeled it’s inner Dionne Warwick and told me Apple had launched an AR headset and a bunch of new Macs
I’ve been listening for the last couple months with AI being talked about on a lot of shows.
The general discussion has been that the LLM is interesting, but not of much value because you can’t really trust the results yet.
Nowhere have I seen anyone mention Peter Theil’s Palentir. This proprietary AI/ML system is about 5 years ahead of the others… already highly in use in Ukraine, and in many companies.
I hear they recently made a deal with Amazon to be made available via AWS to make it easier to integrate into more programs.
Any thoughts on this?
Shawn
You can trust the results from the LLM. models, it’s the sources that cause the problems
Which is why the Bloomberg one fascinates me, as they have generations of their own generated content they can feed it.
Exactly, Bing AI and Bard giver you not much weirder answers than Bing and Google - because it’s néw you look closer at itt