Working Up

by Dwayne Phillips

As much as we “have a better idea” and want to revise, sometimes the first version is the best.

The first draft. The bane of writers of fiction and non-fiction. Let’s fix it. Okay. That’s the first version. Let’s revise as we become “smarter.” Are we always becoming smarter? Sometimes not.

The first draft. That’s the poison pill of writing for many of us. Okay, fix it. Now we have the first version.

In my current job, teams of us are writing non-fiction. We have the first version. We have time before the writing is due. Hmm, I looked at this again and thought of something while driving to the office this morning. It would be better if… Repeat for ten days in a row. Now we are at version ten or twelve or something. Perhaps it is version 20. Anyways, who is counting? Every day brings better thoughts and the document is better and better.

Or is it?

It has to be. We become smarter every day.

Or do we?

Perhaps that is our ego telling us that of course we are smarter every day.

Or are we?

Perhaps. Perhaps not.

Sometimes the first version, written at a time when we are fresh and have clear insight, is the best version. Not the first draft (as I read back through this little five-paragraph piece I find errors) as it needs fixing. After the fixing, the first version is clean, fresh, and it has something innocent about it. There are no patches.

But, you know, just a little bit here and a little bit there and … we often have a patchwork quilt but the blue and the red are from different base colors (I think that is what they call it) and they just don’t go together.

Now what?

Here are two choices; no doubt there are more. (1) Go back to the first version and send it. (2) Go to a blank sheet (or screen or whatever we call it these days) and write from start to finish. Don’t look at the 20th version; don’t copy anything from that version (I hate to type again, surely copying won’t hurt, yes it will) write clean.

I have met writers who have been writing their novel for five years. Do they type slowly? No, they are on version 30 or something. I have met and read from writers who earn their living from writing novels and stories. Writing isn’t a part-time job. It is the full income, and the income is sufficient. They ship the first version and move on to the next piece in which they ship the first version and move on etc.

Sometimes the first version is the best version. That doesn’t seem right to many of us. Nevertheless, it is right. Let’s consider it.

by Dwayne Phillips

I would hope that journalists would understand AI and software better than this. I am disappointed.

Yet another story about AI written by the ignorant for the masses:

AI models that lie and cheat appear to be growing in number with reports of deceptive scheming surging in the last six months, a study into the technology has found.

Ah, yet another story about a programmer who didn’t adequately test the software they wrote. Software “scheming” and “deceiving?” And let’s not forget the “lying.” Good gosh, software that lies about its intentions. Software that has intentions of its own. Good grief.

The software did something the programmer didn’t intend. The board of inquiry, (we really don’t have boards of inquiry to grill programmers, but it sounds neat) asked the programmer questions. “I didn’t write that software to do that!” replies the programmer. Examination of the source code reveals, uh, well, yes, the instructions are there. The programmer needed a DWIM button (Do What I Mean). The programmer didn’t have a DWIM button. The computer performed as instructed.

“Let’s review the test procedure and test report,” says the board of inquiry.

“Well,” says the programmer while crouching lower in his chair in the middle of the room surrounded by the aged, wise, and vicious inquirers (again, we don’t really do this, but it makes for drama in a movie), “I sort of didn’t have a procedure and report, but I did test the software.”

AI models come from programs written by people. People test the programs a lot, a little, or not at all. There are methods of doing these things to prevent these headlines. Let’s do better.

And hey, you journalists and editors out there, do better.

by Dwayne Phillips

Endeavors have fine, precise adjustments. The details matter. There, however, is a time for details and fine, precise, adjustments. The early stages of work is not the time.

I wish I was smarter. I wish I knew all the details and adjustments and the little factors of success at the start of an effort. Alas, wishes are not reality. I am not that smart. I need time to learn. I need space for solutions that come when I know more.

Back in time, a person told me of a project where someone was specifying find adjustments at PDR (preliminary design review, something that occurs early in a project). That was way to early to be fine tuning solutions. There were too many unknowns. There was too much room for learning.

In my current job, I write documents that have strict limits on length. I have five pages for this and two pages for that. I don’t know what I am doing. I will learn as we go along. I draft four pages for this a one-and-a-half pages for that. I leave room for learning. Time will teach me. Then I will fill the blank spaces.

This is humility. I wish I was smart enough to draft a complete document with all the details. Alas, I am not. Leave space. Leave time for learning. We can do better.

by Dwayne Phillips

We have a new twist on the old excuse, “My dog ate my homework.”

That excuse about the dog and the homework goes back to 1905 (some sources report, but then again, there may be earlier instances, but the dog ate the stories and …). John Steinbeck even said that his dog ate the first draft of “Of Mice and Men.” If the excuse was good enough for Steinbeck, well …

Anyways, a student does his homework on a laptop computer. The student later has an AI Agent (we have to use Agents as they are spiffier versions of plain old AI) optimize his laptop computer for some gallant task that deserves optimization. The ever so earnest and eager AI Agent removes the homework file as it is deemed a waste of something or other and therefore a hindrance to optimization.

The honest student goes to the teacher and states quite truthfully, “AI ate my homework.”

The teacher, experienced and dubious, isn’t convinced.

“I’m not making this up,” pleads the student (who doesn’t know how to talk without ending a sentence with a preposition; I digress).

The teacher, knowing the truthfulness of the student, relents and grants an extension to the exercise, ending with, “Print your homework the second you finish it so you have proof in your hands.”

Ah, we return to paper and proof. Let AI just try to reach out of that computer and shred the paper.

Then again, keep the dog away from the paper.

by Dwayne Phillips

We never run out of excuses. Well, if we ever come close to the end of the list, we can add AI.

Microsoft had a mistake. No worry as mistakes can be corrected. The source of the mistake? Well, people are always the source of mistakes, but now we can blame AI.

You see, Copilot (Microsoft’s entry into the list of current AI systems) added some code to something or other and that AI code ran the system off the rails or somewhere unintended. If it weren’t for that darn AI, we wouldn’t have this problem.

Any people involved? Of course there were. Let’s have their names so we can shame them publicly or something like that. No wait, let’s shame Copilot.

Shame on you Copilot!

Hmm, that is satisfying in a way, but sort of hollow. Of course there are nameless people here. We all make mistakes. We try to correct them before something truly bad happens. Why the mistake? Answering, “Because I’m human,” doesn’t seem to quench the thirst of the blamer. Oh, but if we can name something specific like COPILOT DID IT! Well, now that is satisfying.

People write software. People call software to do things. Sometimes people don’t test software enough to find and fix mistakes. We can do better.

by Dwayne Phillips

AI capabilities are increasing fast. Too fast for some of us. What do we call this?

This past week, Anthropic showed people what Mythos could do. WOW! That is amazing. Why only a few months ago… Where were we way back then? And, by the way, how do you pronounce Mythos?

There was this thing called the Claude Bot or Clawed Bot or some clever spelling. Why that was the revolution people wanted. Does anyone remember it? That was in November of last year. Uh, does anyone remember that far back?

I bought a book recently on AI Agents. It’s a good book. Well researched and written and … hopelessly outdated. Unless you write a book in a week and have it on the shelves a day later, it is outdated when the topic is AI. What a shame.

This is all moving fast. That is a gross understatement.

Remember Moore’s law? It described how the density of processors on a chip would double every 18 months. I propose such a law for AI capability. We could call it Moore’s law for AI. I like the idea of calling it Phillips’ Law:

The capability of AI systems doubles every month or sooner in some cases.

Phillips’ Law doesn’t have the ring to it that Moore’s law does. And there is that awkward placement of the apostrophe when your name ends with an “s.” So be it, let’s see if anyone else adopts the title of the law let alone pays attention to the content. This is my little attempt.

by Dwayne Phillips

I would hope that journalists would understand AI and software better than this. I am disappointed.

I recently read this story about a rogue AI Agent:

A rogue AI agent recently triggered a major security alert at Meta Platforms, by taking action without approval that led to the exposure of sensitive company and user data to Meta employees who didn’t have authorization to access the data.

Ooooooo, rogue AI Agents. Here comes the Terminator and the like. Just like the movies. Hide under the bed; flee to the mountains and hide in caves.

Good grief! Someone wrote software, didn’t test it adequately, and things happened that the programmer didn’t intend.

Have you ever written a C program that allocated memory in a recursive function and … uh oh. I didn’t intend to gobble all the memory and lock the supercomputer and have to call the technician to cycle the power to the entire building and call the fire department and police department and … and that didn’t involve any AI.

Come on folks! AI Rogue Agent? It is software someone wrote. “The computer is down.”

Did the AI Rogue Agent delete files? Surely you have backups of the files to recover, right? Oh, you didn’t? Hmm. Well, another lesson. Surely you weren’t connected to the Internet so that everyone’s personal information was pumped out there for all to see? Oh, you were? Hmm. Well, another lesson.

We don’t need AI to do stupid things and suffer the consequences. Let’s do better.

by Dwayne Phillips

Big iron is back. The mainframe is back. And we hate it more now than ever.

A long time ago, in a galaxy far, far away… well it was 1979, Baton Rouge, Louisiana and a friend of my father worked for the computer company of America (IBM). He walked me through the BIG ROOM holding the mainframe computer of the state’s university system. It was a variant of the model 360 computer held in a room that had more square feet than my house. It had 16 MegaBytes of memory. People gasped when we said that number. It was all liquid cooled. How do you cool a computer with cold water?

That was the mainframe computer. That was the computer center. We used the cycles in that mainframe on our class assignments. We didn’t know it, but we were told it was so. The computer center was the dwelling place that few of us ever trod. It was like that place in Petra that Indiana Jones and Sean Connery entered in that movie. It was amazing.

Then some group in Massachusetts (yes the spelling checker got that one right for me) created these PDP-## and VAX-## mini-computers and later some guys in California built a computer using a microprocessor that you put on your kitchen table.

And then we fast forward a few decades and now we have the datacenter or data center (I have yet to resolve that question). Ah! that sacred ground in which few trod, but we are all using the computers inside or at least people tell us we are using the computers inside.

Big iron is back. It is the datacenter. And it is hated by those who hate buildings that occupy acres of land that used to be farms. Funny how those who hate the buildings never worked on a farm yet criticize the grandchildren of the farmers for selling out to big iron and all that stuff. I digress.

The other funny thing is that computers are much smaller and much more powerful than the mainframe of 1979, but we need a bigger building to house them. What happened with all that? I digress again.

Today’s datacenter is the mainframe computer and the computer center of days gone by. We sort of laughed at the mainframe when we saw a computer with an apple sticker on it. Then the logo of the computer company of America appeared on a computer that I could put on the kitchen table. We laughed a bit more. The mainframe went away for a while. The computer center changed its name to data center. What happened to the computer? What is this data stuff?

We awed at the computer center and the mainframe. Then we chuckled at them. Now we disdain the datacenter. Well, at least those of use who weren’t paid all that money for all that land, and concrete, and steel, and copper, and fiber optics, and … the list continues. Since I didn’t get any of that money, I guess I am one of the disdain-ers.

Time moves on. We circle back.

by Dwayne Phillips

We all make mistakes. Some are allowable and some not. There is the mistake budget.

We all make mistakes. The sooner we acknowledge and live that the better we will be (IMHO). Some mistakes, like typographical errors in blog posts, are embarrassing but that’s about the cost. Other mistakes, those that cost a million dollars of taxpayers’ money, are a bit more consequential.

I work with government documents much of the day. The folks who created those documents, the government’s employees, make lots of mistakes. Companies must follow the letter of the law or the letter of those documents. When people see the mistakes, they send questions to the government employees who then send answers. This Q&A costs time and salary money. That money doesn’t just add up, it multiplies. Stating “Friday the 13th” when Friday is the 12th, costs the taxpayers a million dollars. I am not exaggerating; it costs that much.

Now we come to the mistake budget. I learned this from the late author and consultant Jerry Weinberg. He had a few employees. Being people, these employees were mistaken now and then. Jerry would record the mistake and the cost it brought in dollars and cents. As long as the sum of the cost was below a set number (a month’s salary or something), things were fine. If, however, the cost became too high (a subjective but agreed upon number), the employee was no longer an employee as they were unaffordable.

This is all to state that while we all make mistakes, some are allowable and some are not. In government employment, there are no mistake budgets. I think that is a mistake, but that is just my opinion. Note, that there is a tool that we can use or not use. Not using the mistake budget is probably a mistake that itself may be allowable or not. To date, those who manage work in government have decided it is an allowable mistake. Perhaps that may change, but probably not.

by Dwayne Phillips

History repeats itself as the computer can be fully occupied by the efforts of just a few programmers.

There was a time in computing history when there were few programmers. The computers weren’t powerful. A couple of programmers could keep a big computer busy all the time. Then the computers became more powerful. More programmers were needed to keep them busy.

Then the computers became much more powerful, much less expensive, and much smaller. Every person could have their own personal computer. We needed hundreds of millions of programmers to keep all these personal computers busy.

Now we are sort of back to the beginning where the number of programmers needed has shrunk to an alarming level. What happened? I think the really smart programmers put the rest of the programmers into the unemployment line.

A few really smart programmers figured out how to tie a bunch of computers together so they looked like one computer. The Beowolf cluster was one instance of this. There are many others. People let someone else use the CPU cycles of their home computer at night, etc. This sort of falls under Distributed Computing.

Then came the datacenter. I don’t know how many individual computers there are in a datacenter. Probably a million or so. Once you reach a number like that, who knows?

Then some really smart programmer figured out how to run one big program on a million processors in a datacenter. One datacenter—one programmer. One solution to solve all the world’s problems. Well, sort of, but that is a detail.

So we are back to a computer needing only a couple of programmers. Except today, that computer is an entire building full of a million computers. Used to be an entire building filled by one computer. One? A million? what’s the difference (a few zeros).

What’s next? Not sure, but it could be fascinating.