I have been wondering a bit over the news that Apple Intelligence would include linking to ChatGPT and other LLMs from Siri when needed, as opposed to introducing their own-in house AI. With two trillion dollars net worth and a year lead time you'd think they can develop their own LLM model.

After my last post, I think the reason Apple did this is indemnification. They realize AI is shit, but also understand the pressure from their investors and the public to participate in the AI madness.

Clever.

I said it before, but generative LLMs are not to be blindly trusted. I am not just talking about hallucinations, I am talking about understanding a problem, showing you its intermediate logic and then coming to a 90% accurate conclusion. See below for a recent interaction with GPT-4:

Me: Decode the following ciphertext

ChatGPT Sure, I can help with that. Please provide the ciphertext that you would like me to decode.

Me: Pm ol ohk hufaopun jvumpkluaphs av zhf, ol dyval pa pu jpwoly, aoha pz, if zv johunpun aol vykly vm aol slaalyz vm aol hswohila, aoha uva h dvyk jvbsk il thkl vba huk 1.

ChatGPT: The ciphertext appears to be encoded with a simple substitution cipher, likely a Caesar cipher, given the structure of the text. One common method to start decoding such ciphers is by trying out different shifts. Let's apply a common Caesar cipher shift of 7 and see if it decodes into readable text.

Here's how we can decode it:

1. Identify the shift used in the Caesar cipher.
2. Shift each letter in the ciphertext backward by the same amount to get the plaintext.

Let's try shifting backward by 7 positions (since 'Pm' starts at 'C' with a backward shift of 7):

Given the apparent regularity, let's decode the whole ciphertext with a shift of 7:

Pm ol ohk hufaopun jvumpkluaphs av zhf, ol dyval pa pu jpwoly, aoha pz, if zv johunpun aol vykly vm aol slaalyz vm aol hswohila, aoha uva h dvyk jvbsk il thkl vba huk 1.

Decoding using the shift of 7:

If he had explained sufficiently to say, he would be in secret, that is, by so changing the order of the letters of the alphabet, that not a word could be made out and 1.

The problem is, here is ChatGPT's result followed by the actual, correct result:

If there is one thing Apple is known for, is its obsessive pursuit of perfection when it comes to good design. However, I have been finding more and more examples lately of them completely dropping the ball. Here is one example of a terrible UI design choice - look at how low the contrast is between the text and the background when the background is a certain colour. This is just the standard weather background they recently introduced. The text should have toggled to white, like it does on a sunny day.

Try to install the Teams for Work or School app from Microsoft's official web site using their official Edge browser on the latest Windows 11 and you are greeted by:

By many reviews the iRobot Braava Jet m6 mop and the iRobot Roomba s9+ robots are some of the best in their respective classes. 3.9 star review on Amazon for both, and as I said many reviewers rated it top of its class.

Today I want to tell a depressing story of two Bad Robots. See, I have had them now for about 4 months, The mop has completed 48 jobs, 47.5 hours total job time, and surprisingly only 55m² area cleaned according to the iOS app. Of those jobs, it got stuck multiple times, closed various doors on itself (even though I use door stoppers), takes about 2.25 hours to clean 60m^2, leaves streaks most of the time when wet mopping, cannot clean any dirt spots such as from snow boots, cannot pick up any larger pieces of soiling / debris (like rice grains, small pieces of food etc).

The s9 vacuum is worse. If the mop was Mussolini, then the s9 is definitely Stalin. The s9 completed 33 jobs, cleaned 45 hours and emptied the bin 41 times. The s9 will get stuck at least 2 times per job on my Ikea POÄNG chair feet. It will repeatedly get stuck on the same foot within 3 minutes of me rescuing it. It takes an average of 6 hours to vacuum an area of 49m². This is because it has such bad navigation, even though both the mop and the vacuum fully mapped the area on a dedicated mapping run, it constantly follows random and sub-optimal paths ending up draining the battery by driving and not working. So it usually has to recharge 2 - 3 times per job. Granted it has a self charging station but 6 hours? Of that 2.5 hours is charging time, 3.5 is cleaning. Also, when cleaning it does not honour my exclusion zones I defined in the app - to not get stuck on the speaker feet. And it misses many spots as it is just as blind as a bat.