Microsoft expands AI features in Office, but are they any good? Mixed, according to our vulture Microsoft made a number of Office 365 Microsoft 365 announcements yesterday, including touting an expanded set of AI capabilities in its core applications. The new AI features, pulled out alongside a name change for most plans from Office 365 to Microsoft 365, centre around Microsoft Editor, a service which the company claims …
All well and good for admins of the corporate moneybags lot with their swanky Pro Plus* subscriptions, but what about those of us having to make do on (and I use the term very relatively) cheap Business Premium or Business Essentials subs?
* Pro Plus might have some more fully-featured admin tools, but bizarrely costs £2 a month more per user (+VAT) and doesn't include any of Microsoft's cloudy service goodness like SharePoint, Teams or even Exchange.
Is it just me, or does the quote "write polished prose, craft impressive emails, and posts on your favourite sites like LinkedIn, Gmail, Facebook, and more." seem to have an errant comma or three? Perhaps they mean "write polished prose to craft impressive emails and posts on your favourite sites like LinkedIn, Gmail, Facebook and more."
I usually craft any FecesBook post that requires paragraphs in Word first.
Its amazing how many users of it don't understand there's a (post) edit function, but then again the standard of literacy among 95% of the posters, leads me to believe that they couldn't find their asses in broad daylight, with both hands & a map*
*Which they would probably use for toilet paper anyway..
It would depend on your Word settings for Grammar & Refinements (or Grammar & Style) as it was previously known. There is a checkbox for Oxford commas which is possibly connected to your language default (EN-US, EN-UK etc)
Many of the settings have been around since Word for Windows 1.0 nearly 30 years ago (when I started using it) - and labelled as Style rather than Grammar settings. Hasn't stopped endless opinion pieces on Microsoft's grammar rules.
Here here. I was about to make a post very similar to yours although mine was going to have more NOs in it. Having my content be sent to random corporations without my consent is just not ever going to be a happening thing for me. I don't want my editor to have any AI abilities. I don't want my editor to suggest anything at all to me, ever. I don't want my editor to do anything but accept my input and commands, and produce text files on the computer that I'm working on. I can take care of all the other stuff with my brain, which has a lot more to do with producing readable documents than the stuff Microsoft vomits up under the guise of being helpful. Maybe all that nonsense is helpful in maintaining employment for thousands of sub-par Redmond "programmers", but none of it helps me at all.
"Hear, hear", not "here here".
I don't want my editor to suggest anything at all to me, ever
Even with the occasional use of the incorrect homophone, your prose is likely better off for it. I've studied automated proofing tools since Grammatik came out in the mid-1980s, and - much like style guides such as Strunk & White - I firmly believe they do more damage than good. And I've taught college writing (so I've also studied composition theory and rhetoric), so I've seen some bad prose.
At best, these tools reduce personal style and dialectical and individual variation to a bleak, dispiriting, joyless mechanical sludge. Usually they also introduce infelicities incorrectly included in their models, such as false elevation.
There is one royal road to good prose style: Read a lot, and write a lot.
The Editor checks your document not only for basic grammatical and spelling errors, but also for more subjective and complex issues such as inclusivity and gender bias, age bias, wordiness, use of clichés, complex wording, and vague or unnecessary adverbs.
AC: Ramming home points about gender bias?
Meh, MS marketing missed their opportunity of a lifetime there...
It would have sold like hot cakes if they had branded it "pedantic mode"...
But always welcoming a good laugh, very curious how their US style translates in my mother tongue where we have many different ways to say F* off...
Yes, sending corporate information to Microsoft servers? No, not never.
As an aside, I just gave it the Complete Bastard - a document I've been creating over the years with all BOFH episodes in it (1,651 pages or 647,195 words)... It is taking its own sweet time. I'm expecting smoke to start coming out of my laptop in a minute.
Checking for "wordiness, use of clichés, complex wording, and vague or unnecessary adverbs" sounds reasonably close to his six rules for writing. But sending all your writing to the somewhat big brother?
I call it a doubleplusgood Orwellian innovation. (What would Clippy say?)
And that of innumerable other style guides. Few of them are worth reading, much less following.
Richard Ohmann’s “Use Definite, Specific, Concrete Language” is a classic corrective to the prose style guide movement.
English, for all its faults, offers unparalleled riches to writers: its huge vocabulary and ability to incorporate foreign words and phrases without faltering; its vast array of synonyms; its grammatical flexibility; its store of idioms; its accommodation of poetic forms and tropes thanks to its wildly varied orthography and pronunciation; its huge range of dialects and variations. Attempts to deny most of those riches to writers are misbegotten schoolmarmism and should be resisted wholeheartedly. Robotic, cookie-cutter, machine-approved prose does no one any favors.
The cautious approach is to turn off the connected experiences, though it is unlikely that any confidential data is in danger of interception. ....... Tim Anderson 31 Mar 2020 at 19:25
Oh please, Tim Anderson, you cannot be serious. Where the hell have you been all of your life? Cloud Nine?
Golden Rules #001 ....... Any and all confidential data and metadata is in danger of interception for remote officious and unofficial second and third party exploitation and abuse.
And that persistent awkward question always remains to be answered in such circumstances ....... Why such an Epic Present Fail?
And that may be easily answered brutally truthfully with the realisation there be not enough advanced and advancing intelligence in-house for exhibition/presentation/information processing/preprocessing/reprocessing.
Such is an abiding catastrophic systemic vulnerability perfectly ideal for relentless and ravenous exploitation by A.N.Others without such a deficit/learning difficulty/disability/disadvantage/obvious problem and that you can be sure is not ever squandered.
And any and/or all of that can be absolutely terrifying too.
but sure as hell that is how it will be used.
Send a document to anyone and it will be tested against the recipient's pet bias and you will no doubt get reamed. Already imagining companies asking if they can bias score their entire library of literature and documentation into a new low of computer controlled Orwellian new-speak.
We only want creativity at interview time... then you will become robotised …
This sounds like an automated version of a fax conversation I had with a US supplier 25-30 years ago. He criticised everything from my spelling and grammar to my "bizarre date format".
My date format was correct, rather than the backasswards one they use in the USA.
My spelling was correct. I had not randomly substituted Z for S or removed the letter U from words like colour.
My, Oxford educated, English teacher at school taught me not to use what people in the USA laughingly call the "Oxford comma".
I had not phrased things in a manner that would have indicated I was about 6 and had, several times used words of more than 3 syllables.
(I can't remember who the supplier was but I think they made UNIX back up software and were in a "flyover state".)
That conversation was fun but having this discussion with a computer would be tiresome.
"Maybe the challenge of taking someone's mangled, cliché-ridden, gender-biased prose and transforming it into something lucid, balanced and objective is too demanding for AI…"
I had to review many documents written by junior engineers. Many needed a human rewrite before one could leave MS Word to tidy up the details.
It would be interesting to feed this with prose from writers with distinct styles like Hemingway and Henry James(*) and see just how boring the resulting "improved" version is. I suspect the result would be the literary equivalent of Goodness Gracious Me's "Going for an English" - I want it really bland.
(*) Or for that matter, Barbara Cartland and EL James.
Just wanted to ask an AI type question. If yoy could preserve state of mind after death, could you make it still develop opinions?
Mine change every day, so just copying my brain would only give a snapshot.
I guess I'm really asking, what is an opinion, and can a machine have one?
Maybe it's just a random result based on incomplete ata available, to try and answer my own question.
You're asking "once we invent a computer that can simulate a human, how will it work?" So the answer is probably "something something quantum something something uncertainty principle". I'll refrain from adding "blockchain" and other buzzwords of equal predictive value.
But an opinion is mostly reasoning from uncertainty. "I don't actually know all the stats, but clearly your sports team is pants and mine is superior." A proper algorithm would define what makes a team "good" (marketing budget? beer endorsements?), then compare both teams on that metric. When we don't agree on what metric we're using or the actual data to measure it, but want to assert something anyway, that's an opinion.
If we look at neural networks now, they need concrete input, and we tell them which results we think are right. But exactly what data and rules they used to get the "right" answer are usually unknown. "We excluded race from our data, so it can't be a factor." Except that postal code, level of education, given name, and who knows what else may have subtle to strong correlations, and from the aggregate of those, we could probably ask a neural network to guess race, household income, level of education, etc. with a much higher degree of accuracy than any one input factor. But instead we ask it to make a lending decision, and for all we know, one of the "hidden layers" is actually deriving race and using it in the decision. One could call that bias an opinion.
Putting a neural network back into learning mode can change these biases, so perhaps that's "still develop[ing] opinions". But it's really just manifesting the biases inherent to its training data, so the question is whether you accept that humans are just doing the same thing when we form opinions.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
