Elephants are some of the most intellingent mammals, and they have large brain sizes compared to people. They are said to present self-awareness among other things.
How could one go about raising the intelligence of elephants to a level closer to humans? I imagine a thousand year project doing this.
Force them to solve problems, of course. Elephants are not smarter than they are because they don't need to be. Humans and carnivores are relatively smart because hunting requires more intelligence than eating stationary plants (for the most part). Many animals can also learn by mimicry, so teaching a few elephants to do cool tricks would also let younger generations bootstrap their knowledge by observation. So...how do you force an elephant to solve a problem?
This is a mild spoiler for a decent story. But the short answer is: pain. You give them a puzzle that they must solve, and longer they fail to solve it, the longer they are exposed to pain. You can also force them to communicate by the same means: give one confined elephant information, and put another in pain, which they can only escape when the first elephant communicates the information to the other. In fact, we know this technique works, because we already have elephants who are smarter than wild elephants: circus elephants. And the way you train a circus elephant is via pain (aka bull hooks).
You can accelerate their learning by giving them an explicit language. You can either construct a language from scratch, and use the pain technique to force them to learn it, or you can try to associate existing vocalizations and movements with particular symbols, imbuing their natural movements with semantic meaning. For instance, you can teach elephants a sign language by showing them images, moving their trunks and ears in a particular way to demonstrate the sign associated with the image, and use positive and negative reinforcement to seal the association. You can also make them learn to recognize signs by a similar system: one elephant is shown a picture, then "encouraged" to sign it to another elephant, who then must select the correct picture from the other side.
I think these techniques alone would get you quite far even in one generation. Selective breeding of the best learners would probably yield promising results, especially if one could augment with any kind of dietary supplements which favor higher brain utilization. In fact, changing their diet to one which provides more glucose for brain activity would also help (one theory is that human brains could expand when they mastered fire for cooking, unlocking more useful calories in food).
If you have advanced technology, you may be able to implant chips directly into their brain to stimulate their visual cortices, etc. This would accelerate their ability to communicate, as well as the ability for their teachers to communicate with them. Once you teach them an elephant language, you can teach them human ones, and then they can learn by [forced] reading. Instead of spending all day eating plants, they could spend all day learning. Surely you would reach the limits of their neural capacity in short order!
With current technology we can't. We don't have enough knowledge about what it means to be intelligent to even begin to make such a thing happen.
If we were going to uplift Elephants using some form of future tech handwavium we'd do it by making them more intelligent. Probably by tweaking their genetics in such a way that they have a more intelligent brain.
Pick some cool sounding brain locations (frontal lobe), pick a cool sounding method of genetically modifying an organism (CRISPR), munge them together in a cool sounding sentence ("We used CRISPR to enhance their frontal lobes and ever since they've demanded that we give them the same rights as people.") and you have an excellent description of how your super smart elephants were uplifted.
Sheer brain size is not very important, encephalization quotient is considered more representative and elephant (1.2..2.0) one is only marginally superior to cat (1.0..1.8) and much lower than Homo Sapiens (7.2..8.0).
This method is not universally accepted, mainly because it gives only a rough estimate. Main idea behind this is a certain (large) amount of the brain is used to control and "model" the body, i.e.: it's used for "general maintenance". Since nervous tissue is very expensive, in terms of nutrition requirements, there's a strong evolutionary push to keep it to the "minimum required". A much better approach would be to really try to understand exact functions of the various brain structures. This has been attempted, but attempts have been hampered because, while certain functions are quite clearly localized, other (especially "higher" functions like language) are only pertly localized (even lateralization is far from being clear-cut), many functions are carried out involving large sections of the brain, sometimes virtually all of it. In practice encephalization quotient, corrected with any known "special requirement" (e.g.: dolphins "excess cerebral mass" is largely used to implement their sonar sense) gives a reasonable hint without the obvious anomalies the sheer mass has (sperm whale should be much smarter than we are and women should be less intelligent than males (yes, I know a lot of people is convinced of the latter, but that doesn't make it true!)).
The key to reach self-awareness is language. to date we are the only species who developed one (not a set of signals, but a real language capable of grammatical and syntactic transformations).
I need to expand this to explain I consider a language a construct following grammatical and syntactic rules capable of recursive definitions, so that is able to describe and manipulate concepts not immediately related to physical objects. Regardless of "common usage" I do not consider the simple passage of information as sign of existence of a real language, otherwise bees would have a language (fairly structured too, but inflexible conveying information on a single topic) as most, if not all, animals.
Language creation seems to be a necessity of the human species, when in sizable communities (see: Nicaraguan Sign Language), while humans grown without contacts with other people won't develop one and can't really be taught a language anymore (there have been a few cases of children raised by animals in the wild). This characteristic seems to be what put us apart from other animals and strictly connected with what we call "self consciousness".
This process could be duplicated, of course, if we knew what really makes the difference, but I strongly suspect we still lack the necessary insight to understand what needs to be done.
Fund big studies on intelligence in rats.
Once we have some clues about how intelligence works in rats then try similar techniques with elephants.
Rats are better because they breed really fast even in captivity and we have less compunction about killing them.
Selective breeding (or culling) might in time raise the average, but in an uplift you care more about peak than average, and it has not been proved to advance the peak. Wait until we have figured out intelligence in rats to try it on elephants seems like a much more likely path.
You can probably selectively breed them for your definition of intelligence. 1000 years however isn't really long enough to get good results because of how long each generation takes for an elephant.
We have been breeding dogs for tens of thousands of years and there is definitely some selective pressure there from humans to get more intelligent dogs. People would want dogs that are social to humans and are easy to train. Dogs will probably score better than wolves at human intelligence tests because they been bred to have all the human characteristic enhanced.
The problem however, with breeding elephants with selective breeding is how long it takes to breed them. A dog has a life cycle of ~15 years and can reproduce at age 5, which means a generation of dogs to the next is ~5 years. For elephants, females are fertile around 10 while males take about 15 years. So each generation takes more than 2x as long. In 1000 years, you can only do about <100 generation of elephants and that is just too few to selectively bring out traits you want. Keep in mind even with tens of thousands of years of selective pressure for smarter dogs, they are still nowhere near what most people would consider human intelligence.
You might be able to breed "smarter" elephants in that time but it is unlikely they'd show the difference you are after.
Rather than only using genetic modification (either artificial or through selective breeding) consider using digital enhancement. Interfacing computer circuitry into brains could allow for a number of modifications and improvements in cognitive abilities much faster than breeding elephants.
This would also provide means for the elephant cyborgs to communicate and interact with their surroundings without a large amount of changes to their physical form. They could communicate with a speaker/microphone based speaking system or through other digital communication methods and they could use robotic arms or other manipulators to allow more dexterous tool usage (although elephant trunks are actually quite dexterous as is).
