May 16, 2017 at 11:30 am
This is just me, and maybe some will disagree. Creating algorithms that operate within a set of predefined rules is not all that complicated unless those rules become extremely complex and factor in large amounts of variables. Thus, creating a machine to always make the best move within the board is not going to be as complex as creating the optimal board itself to me.
Humans will still be used to create the board and expertise from those such as DBA's will still be critical in that board creation even before we get into the pieces and optimial movements of those pieces.
The only question and fear I would have is at what point do we let the machines create the board too? When do we allow the machines to procreate? At that point, this may be Skynet and we are all enslaved by Terminators!!! :laugh:
May 16, 2017 at 11:17 pm
xsevensinzx - Tuesday, May 16, 2017 11:30 AMThe only question and fear I would have is at what point do we let the machines create the board too? When do we allow the machines to procreate? At that point, this may be Skynet and we are all enslaved by Terminators!!! :laugh:
You take the words out of my mouth! I just hope we don't go so pro-machine that iRobot and Terminators come true!!
Manie Verster
Developer
Johannesburg
South Africa
I am happy because I choose to be happy.
I just love my job!!!
May 16, 2017 at 11:48 pm
Many companies are shifting their DBs to the cloud. Backup, high availability, monitoring and alerting are all options that are not much more than check box items.
The hardcore DBA skills just vanish in such a case except for what the cloud providers need. In many cases you don't have permissions to do what you as a DBA would have done and with that reduced permission set goes some of the queries you use to dig deep into performance problems. It can be tremendously frustrating but at the same time greatly liberating.
There are very few cases which cannot be moved to the cloud. The "We're a bank it will never happen" argument doesn't stack up. AWS worked with one of the banks to get full FCA (UK) approval. An AWS Architect can help design a system that is probably more secure than most on-premise systems.
I would say that the analogy of humans designing the board before computers can play the game was a good one.
The bit that a machine cannot yet do is the random "I wonder if..." that uncovers an unexpected beneficial behaviour or feature
May 17, 2017 at 6:48 am
David.Poole - Tuesday, May 16, 2017 11:48 PMMany companies are shifting their DBs to the cloud. Backup, high availability, monitoring and alerting are all options that are not much more than check box items.
The hardcore DBA skills just vanish in such a case except for what the cloud providers need. In many cases you don't have permissions to do what you as a DBA would have done and with that reduced permission set goes some of the queries you use to dig deep into performance problems. It can be tremendously frustrating but at the same time greatly liberating.There are very few cases which cannot be moved to the cloud. The "We're a bank it will never happen" argument doesn't stack up. AWS worked with one of the banks to get full FCA (UK) approval. An AWS Architect can help design a system that is probably more secure than most on-premise systems.
I would say that the analogy of humans designing the board before computers can play the game was a good one.
The bit that a machine cannot yet do is the random "I wonder if..." that uncovers an unexpected beneficial behaviour or feature
I kind of view the cloud as a better option and placement for some instances. For example, large organizations may have centralized IT teams where you may have local resources dedicated to specific products. The DBA may be limited in those instances because they are tied locally to the product versus globally with the organizational operations. Movement of the cloud may allow that same DBA more control over a specific environment where they gain more insights into the entire operational health and management of the specific applications they control. It's just a matter of the fact that most of the operational pieces are taken care of the cloud provider, but at least you still have great insight into what they are doing in terms of those backups and so forth or even change them.
May 17, 2017 at 7:17 am
David.Poole - Tuesday, May 16, 2017 11:48 PMMany companies are shifting their DBs to the cloud. Backup, high availability, monitoring and alerting are all options that are not much more than check box items.
The hardcore DBA skills just vanish in such a case except for what the cloud providers need. In many cases you don't have permissions to do what you as a DBA would have done and with that reduced permission set goes some of the queries you use to dig deep into performance problems. It can be tremendously frustrating but at the same time greatly liberating.
...
Honestly, I've always felt that hardcore and routine DBA tasks are simply mundane and a distraction from other DBA tasks that I'd rather be doing. For smaller organizations where the DBA team isn't specialized; the single jack-of-all-trades DBA may find cloud hosted database services to be liberating. Once upon a time application programmers were system engineers; they had to know the internals of PC hardware, the operating system API, and machine set instructions just to do their job. All that changed with the invention of high level programming languages and GUIs. It's essentially the same thing.
"Do not seek to follow in the footsteps of the wise. Instead, seek what they sought." - Matsuo Basho
May 19, 2017 at 10:45 am
Eric M Russell - Tuesday, May 16, 2017 7:36 AMJeopardy or Chess don't come close to modeling anyone's actual day job; unless you're a professional game player. Perhaps one way of testing the theory (that an AI could replace a human DBA) is for Microsoft to create a Virtual SQL Server MVP and have it participate in discussion forums. See how it scores over time compared to human counterparts.
Only if you try to take your day job as a total package, meaning some random set of things that people ask for you to finish. If you look at your job as a series of individual items, then a number of those can be replaced, especially index tuning, managing jobs and more.
It's not that an AI eliminates all DBAs. It can provide huge leverage to a few DBAs, so there are many, many less DBAs needed at scale.
May 20, 2017 at 7:22 am
The human brain is extremely powerful regardless. We can factor in so many things even if you don't think you're all that smart. We can quickly look at something and make a decision on what we want to do: pick it up, throw it, eat it, walk away, talk and walk away, etc. These are all things we are calculating in our minds even to a point of what seems to be without thought. Just taking driving. I barely notice me thinking, but I am and I just do it naturally. Even self-driving cars are not there yet and we've been doing it for, well you know.
In my opinion, to start down the path of replacing the board with AI, you're going to need to develop something like the creator of the matrix would have to do. Build the entire world from top to bottom in order for AI to make decisions. Realistically, this is going to be simulated by advanced algorithms that will allow AI to build the scene, build the rules, and build the actions to react to them as opposed to a engineer defining the entire world in it's memory.
We are ways away from that I feel unless I'm totally ignorant to some brave new technology that can totally replace a human being in terms of the human brain. In the meantime, smaller tasks will be able to be automated and replaced by AI/and or machines. Backups, index rebuilds, statistic updates, checksums, performance tuning, audits, etc. In other spaces like retail, this could be customer service, POS, store cleaning, and other really simplistic tasks that can be easily programmed.
May 21, 2017 at 2:08 pm
xsevensinzx - Saturday, May 20, 2017 7:22 AMThe human brain is extremely powerful regardless. We can factor in so many things even if you don't think you're all that smart. We can quickly look at something and make a decision on what we want to do: pick it up, throw it, eat it, walk away, talk and walk away, etc. These are all things we are calculating in our minds even to a point of what seems to be without thought. Just taking driving. I barely notice me thinking, but I am and I just do it naturally. Even self-driving cars are not there yet and we've been doing it for, well you know.In my opinion, to start down the path of replacing the board with AI, you're going to need to develop something like the creator of the matrix would have to do. Build the entire world from top to bottom in order for AI to make decisions. Realistically, this is going to be simulated by advanced algorithms that will allow AI to build the scene, build the rules, and build the actions to react to them as opposed to a engineer defining the entire world in it's memory.
We are ways away from that I feel unless I'm totally ignorant to some brave new technology that can totally replace a human being in terms of the human brain. In the meantime, smaller tasks will be able to be automated and replaced by AI/and or machines. Backups, index rebuilds, statistic updates, checksums, performance tuning, audits, etc. In other spaces like retail, this could be customer service, POS, store cleaning, and other really simplistic tasks that can be easily programmed.
True, if you're trying for 100% replacement There are many tasks, including pieces of driving a car, that a computer and an ML/AI type system does way, way better than humans. They don't do everything better, and they certainly aren't creative, but when we have fairly set rules, the models tend to outperform humans most of the time.
Keep in mind that as well as humans can drive, they also fail at driving quite often. We don't need perfect AIs that make every decision, or replace all humans. They only need to augment and amplify human capabilities, allow one human to do what 10 did before, to be incredibly disruptive.
May 21, 2017 at 4:18 pm
Steve Jones - SSC Editor - Friday, May 19, 2017 10:45 AMEric M Russell - Tuesday, May 16, 2017 7:36 AMJeopardy or Chess don't come close to modeling anyone's actual day job; unless you're a professional game player. Perhaps one way of testing the theory (that an AI could replace a human DBA) is for Microsoft to create a Virtual SQL Server MVP and have it participate in discussion forums. See how it scores over time compared to human counterparts.Only if you try to take your day job as a total package, meaning some random set of things that people ask for you to finish. If you look at your job as a series of individual items, then a number of those can be replaced, especially index tuning, managing jobs and more.
It's not that an AI eliminates all DBAs. It can provide huge leverage to a few DBAs, so there are many, many less DBAs needed at scale.
Very true. But that's not AI. I'm not even sure that qualifies as being heuristic.
--Jeff Moden
Change is inevitable... Change for the better is not.
May 21, 2017 at 7:57 pm
There are certainly places where AI like systems have replaced humans and are used for controlling items. Parts of Japanese trains are run by AI systems, advanced neural nets that keep things on time. There are AI systems being tested by MS to handle indexing and make good choices based on workload in Azure.
AI will impact us and will come in various forms. If you are expecting AI to only be some system that has sentience and can replace a human in many, or most, things, then I'll admit that's not likely to happen anytime soon. If you see an AI system as one that isn't directly programmed and grows or learns over time, then that's going to happen.
May 21, 2017 at 11:43 pm
I would argue you need to have failure too. Anyone remember what happened to Joshua?
I'm a yougin, so hopefully you older guys pick up on that one muaha :hehe:
May 21, 2017 at 11:44 pm
Steve Jones - SSC Editor - Sunday, May 21, 2017 7:57 PMThere are certainly places where AI like systems have replaced humans and are used for controlling items. Parts of Japanese trains are run by AI systems, advanced neural nets that keep things on time. There are AI systems being tested by MS to handle indexing and make good choices based on workload in Azure.AI will impact us and will come in various forms. If you are expecting AI to only be some system that has sentience and can replace a human in many, or most, things, then I'll admit that's not likely to happen anytime soon. If you see an AI system as one that isn't directly programmed and grows or learns over time, then that's going to happen.
There's no doubt that there are some very clever programs running on some very fast machines. There is no question that machines can learn from their inputs thanks to such clever programming and long term storage of data. But if you define AI as those things that you've described, then I'm disappointed in the term AI.
I've written some T-SQL that can "look" at a spreadsheet to determine its "shape" and auto-magically import the data deciding on the way which data should be unpivoted as data and which data should be used as identifiers whether they be horizontal headers or vertical in a mix of both. Next month, the spreadsheet may have suffered a great number of additions of rows and columns along with other substantial "shape" changes including the type of headers it may be sporting. It requires no human interaction and has replaced the requirement for a human to map the spreadsheet columns to table columns and has replaced the requirement for a human to decide which headers belong to common elements and should be used as identifiers and which columns to apply as data to those column identifiers. The result table also changes "shape" to accommodate the different dimensions found in the spreadsheets. In other words, it's adaptive to the varying problems that it's been given much like a human would need to be to solve the same problem. It even writes it's own code once it has determined the "shape" and locations of items on the spreadsheets. If I were to add a bit of code to "train it" to remember what it has done previously, it could improve itself by trying learned "shape" tests much like the T-SQL Optimizer does although I believe that might actually slow it's decision making down. Since some of the tools that it uses are mutually exclusive when it come to concurrency, my next step would be to make it realize that other instances of the proc are running and that it should wait and try again later. The same goes for if the user currently has the spreadsheet open. One more tweak and it will be able to decide if the spreadsheet looks like a "normalized" table and can be imported without any mapping or if it looks like something that will need to be mapped and unpivoted.
Considering all that it currently does and will do and the fact that it replaces all human effort in the area, is that "AI" according to your or anyone else's definition? I don't believe so because I trained myself to identify the "shape" of a spreadsheet whereas the proc had to be trained to do such things.. I think that it's "just" a bit of clever programming just as I believe that a chess playing computer or self driving car is. Don't get me wrong. Those are tremendous computational successes and my hat is off to the people that made such things come to be but... they're not artificial intelligence in my book.
Shifting gears back to the subject at hand and whether or not we agree on the definition of AI doesn't change the rest of what you said, though. Systems can learn and grow based on their inputs and measured outcomes... if they are programmed to do so. Such programs can, indeed, make certain aspects of being a DBA obsolete. If you stop and think about it, SQL Server made certain types of programming virtually obsolete. For example, instead of having to figure out the shape of a part of a file, where all the pieces are, which rows to return, how the columns are structured, and then write loops to go through all of that to return the data I want, I just type SELECT and a couple of hints as to what I want done and it figures out how to return the desired result set automatically. I don't consider even that to be AI.
--Jeff Moden
Change is inevitable... Change for the better is not.
May 22, 2017 at 12:01 am
Jeff, if you want a pay rise then yes it's AI and probably as credible as a lot of stuff that salesmen are going to be peddling. In much the same way as many Big Data solutions really weren't Big Data (or solutions) just a mildly tweaked existing product with last years "Cloud Enabled" starburst sticker peeled off.
Its going to be hard to identify the 2% genuine and good solutions vs the snake oil
May 22, 2017 at 8:35 am
Automated customer support is an example of one application where AI really shines. No?
:rolleyes:
"Do not seek to follow in the footsteps of the wise. Instead, seek what they sought." - Matsuo Basho
May 22, 2017 at 9:24 am
David.Poole - Monday, May 15, 2017 3:46 PMsmayer - Monday, May 15, 2017 12:57 PMOne thing no level of AI can ever beat -- micro-management (i.e., lack of trust). This can be, and often is, a DBA's worst nightmare, and must be rooted out before AI can ever replace a DBA.Are you saying that DBAs need to learn to trust AI?
No, but rather managers need to trust their DBA's first, before AI could even be considered.
Viewing 15 posts - 16 through 30 (of 34 total)
You must be logged in to reply to this topic. Login to reply