November 2022 might not feel like that long ago, but in the artificial intelligence world it’s an eternity.
When ChatGPT–the vehicle for OpenAI’s tour-de-force large language model (LLM)–was unleashed just last year, it seemed almost impossibly advanced. Now Bard, Google’s LLM response to OpenAI’s pioneering AI technology, is catching up. The question is, how fast? How does it compare? And how can it be used for work in higher education?
Bard's Potential in Higher Education
If you’re in higher education, AI has caused its fair share of hand-wringing, but it’s also brought a new level of efficiency that many in the industry have come to embrace. Yes, it’s difficult to create assignments that don’t lend themselves to easy cheating, but it’s not like professors aren’t taking advantage of these new tools. There’s even an online community of teachers devoted to figuring out how ChatGPT can grade their essays.
In a recent survey, 97% of teachers claim to use ChatGPT to write lesson plans, and 93% use it to grade and give feedback. Basically, AI is here to stay–especially now that Bard has entered the race.
Bard as a Research Tool for K-12 and Higher Education
Bard draws on the most relevant information and makes real attempts at transparency. Current LLMs hallucinate facts, so Google provides context for its AI-generated content through source citations, photos, and the option to “Google it”. For researchers that’s a huge plus. If something seems iffy, it makes the fact-checking process a lot easier.
Bard also has the most up-to-date information. It draws from Google search results, so it stays current, while ChatGPT is only trained on info up to 2021. You could pay $20 a month for GPT-4/ChatGPT Plus, which stays up-to-date with Bing, but why bother when Bard does the same for free? (Also, who Bings??)
Bard’s interface also delivers a more user-friendly research experience. Here’s a quick rundown:
Chunks text for better readability
Uses bullet points effectively to make things easier to skim
Automatically includes pictures for sources and context
Offers alternative drafts that present info differently
When it comes to Bard’s voice, its language is somewhat spare and to-the-point. You get the sense that it’s primarily concerned with providing an objective summary that covers the key details. For people doing research in higher education, or teaching their students to do the same, it’s a user-friendly, expedient portal to more information. The elevated context it provides also helps instructors and students to quickly evaluate whether that information is correct–which is a key skill in this new AI-driven research landscape.
Bard as a Writing Tool for Higher Education
Bard’s style is great for objective summaries, but it tends to be stilted when you want to use a more nuanced voice and tone. Take this email for instance, where I ask Bard to write an apology to a student for a grading error.
It’s a little generic and perfunctory–especially when compared to how thick ChatGPT laid it on the very first try…
Bard got somewhat more heartfelt when I asked it to, but it hardly seemed to convey any real sense of humanity. ChatGPT strives for that humanity–sometimes to the point of verbosity and redundancy. Still, it gets points for trying to capture that human touch people crave in their interactions with tech.
As a writing tool, this limited creative capacity causes Bard to falter where ChatGPT flourishes. For instance, if you want an in-depth breakdown of how Sigmund Freud would analyze the character, Jack, from the show Lost through the lens of relational dynamics, ChatGPT steps up to the plate:
ChatGPT on Jack's Characterization
“Analyzing Sigmund Freud’s theories in relation to the character Jack from the TV show “Lost,” we can explore potential aspects that Freud might have considered:
Father Complex: Freud’s Oedipus complex theory suggests that individuals may have unconscious desires and conflicts in their relationships with their parents. Considering Jack’s complex relationship with his father, Christian Shephard, Freud might examine any unresolved father-related conflicts or desires that could influence Jack’s actions and decisions throughout the series.
Defense Mechanisms: Freud proposed defense mechanisms as psychological strategies individuals use to protect themselves from anxiety or discomfort. Jack displays various defense mechanisms throughout the show, such as denial, repression, and displacement. Freud might analyze how these defense mechanisms relate to Jack’s struggles with his identity, relationships, and the challenges he faces on the island.
Ego and Superego: Freud’s structural model of the mind includes the ego and superego. Jack, as a leader and a man driven by a sense of responsibility, might represent the ego and superego in the context of the island’s group dynamics. Freud might explore how Jack’s ego mediates between the conflicting demands of survival, morality, and the well-being of the group.
Self-Identity and Self-Exploration: Freud’s work on identity and self-exploration could be applied to Jack’s character arc. Jack goes through a journey of self-discovery and identity formation on the island. Freud might analyze how Jack’s experiences and relationships contribute to his exploration of self, the resolution of internal conflicts, and the development of his sense of purpose.
Death Instinct: Freud introduced the concept of the death instinct, which represents an unconscious desire for self-destruction or the return to a state of non-existence. Given Jack’s self-sacrificial tendencies and his pursuit of purpose, Freud might explore how the death instinct manifests in Jack’s actions and motivations throughout the series.
It’s important to note that these interpretations are speculative and subjective. Freud’s theories are subject to different interpretations and have limitations in their application to fictional characters. The analysis of Jack’s character from a Freudian perspective would be hypothetical, and actual interpretations may differ based on individual perspectives and understandings of Freud’s theories.”
Now that's a character analysis. Elevated, sophisticated, and comprehensive. Bard, on the other hand, gives a more facile explanation...
Bard on Jack's Characterization
"Sigmund Freud would likely analyze Jack Shephard, the main character of the TV show Lost, as a classic example of an ego personality type. Ego personalities are driven by a need to be in control and to be seen as competent and capable. They are often seen as leaders and are typically very successful in their careers. However, they can also be seen as controlling and demanding.
In Jack’s case, his ego personality is evident in his need to be in control of the survivors of Oceanic Flight 815. He is constantly making decisions and trying to keep everyone safe. He also has a very strong sense of justice and often takes on the role of the moral compass for the group.
Freud would likely see Jack’s need to be in control as a way of compensating for his underlying feelings of insecurity. Jack’s father was an alcoholic who was often absent from his life. This likely left Jack feeling like he had to be the strong one and take care of everyone. As a result, he developed a strong ego personality in order to protect himself from feeling vulnerable.
Freud would also likely see Jack’s relationship with Kate as an example of his ego personality. Jack is constantly trying to control Kate and make decisions for her. He also has a very possessive attitude towards her. This is likely because Jack sees Kate as a way of filling the void left by his absent father. He needs to feel like he is in control of her in order to feel secure.
Overall, Freud would likely see Jack Shephard as a classic example of an ego personality. Jack’s need to be in control is a way of compensating for his underlying feelings of insecurity. His relationship with Kate is also an example of his ego personality. Jack needs to feel like he is in control of Kate in order to feel secure.”
Bard isn’t necessarily wrong. It’s just not enough to satisfy the Lost fanboy in (hopefully) all of us; and leaves you with the fairly common Bard aftertaste that it’s impressive but…unfulfilling.
That might be because it’s based on, well, Google. It searches the internet like Google does, so it finds the popular recent stuff on a subject. Once it finds the information that seems to meet the request, it stops. That’s all well and good, but compared to the depth of ChatGPT’s info-splurges it often looks incomplete.
This would be absolutely forgivable if you knew you were sacrificing volume for accuracy, but that’s not always the case.
Bard Accuracy vs. ChatGPT Accuracy
You'd think that Bard, with its Google-Powered info updates, would own the accuracy conversation, but it's kind of a mixed bag. Bard pulls more relevant recent information, which makes it useful (even though ChatGPT Plus can do that, too). However, that recent information can be misleading.
LLMs can’t actually tell the difference between truth and falsehood. They’re really just highly efficient word predictors that can use training and context to predict what the next word or phrase will be. That’s really amazing tech, but if you let it loose on the internet, where sometimes it’s difficult to parse what’s true and false, there can be issues.
This is especially true with new topics or events, where there’s less context to provide a strong sense of the truth. Bard has a tendency to hallucinate when considering recent topics or subjects that haven’t been widely covered.
For instance, when I asked Bard about Halda (an education website personalization company) and described us as a tech company, it could give me a solid description of the company and our location, but totally whiffed on our logo and who our clients are.
Even with older, well-covered subjects, Bard can be inconsistent. Higher ed marketers often use LLMs as campaign development tools, so I asked bard to give me some famous taglines that could work for university slogans.
It did give me some taglines, but they weren’t from Harvard and Stanford, as Bard claimed them to be. And the output was quite limited compared to the large and detailed list provided by ChatGPT.
K-12 and Higher Ed Marketing: B
Bard can do all the stuff that ChatGPT can do. It can change tones, write emails and blogs, brainstorm ideas, provide feedback on your work, help you do research, etc., it just does most of those things worse than ChatGPT. That’s a big reason why Halda incorporates ChatGPT’s generative AI into our own personalization products. It simply provides a more authentic human experience.
When it comes to staying up-to-date, however, Bard definitely provides some legit help. Marketers can use it to research new trends and directions in a constantly changing industry.
So what does this all mean for Bard’s effectiveness in K-12 and higher ed marketing? Great research tool…limited content-generation tool.
Bard and Lesson Planning: B
Bard’s relative lack of depth and focus on detail make lesson planning with the tool a bit of a give and take. I asked both products to put together an introductory lesson plan for the novella Heart of Darkness. Bard’s first attempt was not specific to introducing the book, while ChatGPT provided a meaningful introduction without any special prompting (although both need more training on what constitutes a theme).
Bard could give me something usable for introducing the book, but only after some prompting–and it never came close to ChatGPT’s first try.
Bard and Academic Research/Writing: B
Considering Bard’s primary role as a research tool, it should give ChatGPT a run for its money if, say, you wanted to track down seminal texts to include in a literature review on Agenda-Setting Theory. In this area, Bart did, indeed, come through. When put to the test, both came up with acceptable lists of texts that set a great foundation for research, with suggestions on how to proceed and handy summaries of each piece.
However, from an academic writing standpoint, while it’s great for quick summaries and basic ideation, it’s a long way from publication. Academic work and teaching requires a level of sophistication that Bard just isn’t yet designed to deliver, especially when compared to ChatGPT.
The Verdict: ChatGPT Prevails
If Bard had come out last November instead of ChatGPT, it would probably get a solid A and we’d all be freaking out. Instead, it gets graded on a curve. In so many ways it’s remarkable, but in many of those ways it’s slightly less remarkable than ChatGPT. That being said, its up-to-date info, user-friendly interface, objective style, and readability make it a useful AI tool.
If you want to continue the conversation about how you could use these different types of AI in your education role, check in with Halda. We're always happy to provide some insight.